Spaces:
Running
Running
File size: 60,086 Bytes
8a0ccd0 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 1d5c0fa 9d2adc1 90b50bc 9d2adc1 8a0ccd0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 | <!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Resume Builder Agent — evidence-grounded resumes from GitHub</title>
<meta name="description" content="Evidence-grounded resumes generated from GitHub repositories. Never invents an achievement or a number.">
<link rel="stylesheet" href="style.css">
</head>
<body>
<div class="wrap">
<header>
<h1>Resume Builder Agent</h1>
<p class="lede">Reads a GitHub account, turns each repository into
evidence-backed claims, and builds role-targeted resumes. It never invents an
achievement or a number — anything it cannot trace to a repository is
dropped and reappears as a <code>[MEASURE: …]</code> task.</p>
</header>
<div class="cards"><div class="card"><div class="n">102/111</div><div class="k">claims verified</div></div><div class="card"><div class="n">6</div><div class="k">claims refused</div></div><div class="card"><div class="n">57</div><div class="k">public repos scanned</div></div><div class="card"><div class="n">21</div><div class="k">repos in this report</div></div><div class="card"><div class="n">11,359</div><div class="k">Groq tokens for this run</div></div></div>
<div class="note">
<strong>This page is a published artefact, not a live app.</strong>
Gradio and Docker Spaces require a paid plan on this account — the Hub
returns <code>402</code> with <em>“hosting Gradio and Docker Spaces on free
cpu-basic requires a PRO subscription”</em> — and Static Spaces serve
files rather than Python. Everything below was produced by a real run; the
<a href="#trace">run trace</a> records what it cost and what it refused to say.
To run it yourself, see
<a href="https://github.com/aniketqw/resume-builder-agent">the repository</a>.
</div>
<div class="note">
<strong>Public repositories only.</strong> This sample was generated from
21 <em>public</em> repositories. The account also has private ones; they
are filtered out before anything is fetched, and
<code>scripts/check_sample.py</code> fails the build if a private repository
name reaches this page.
</div>
<div class="tabs" role="tablist">
<button role="tab" aria-selected="true" aria-controls="p-report" id="t-report">Sample report</button>
<button role="tab" aria-selected="false" aria-controls="p-trace" id="t-trace">Run trace</button>
<button role="tab" aria-selected="false" aria-controls="p-how" id="t-how">How it works</button>
</div>
<section class="panel report" id="p-report" role="tabpanel" aria-labelledby="t-report">
<blockquote>
<p><strong>Public sample.</strong> This report was generated from 21 <strong>public</strong> repositories
belonging to <code>aniketqw</code>. Private repositories are excluded by construction —
see <code>scripts/check_sample.py</code>, which fails the build if a private repository
name appears here.</p>
</blockquote>
<h1 id="resume-report-aniketqw">Resume report — aniketqw</h1>
<p>Built from 21 repositories. <strong>102 claims verified</strong>, 3 need measurement, 6 rejected.</p>
<h2 id="security-findings-in-your-repositories">Security findings in your repositories</h2>
<p>These were found while reading your repositories. Fix these first.</p>
<ul>
<li><strong>[high]</strong> <code>pehenava-backend/pehnava-ai-frontend/.env.local</code> — An .env file is committed. These normally hold live credentials.</li>
</ul>
<h2 id="role-comparison">Role comparison</h2>
<div class="scroll"><table>
<thead>
<tr>
<th>Role</th>
<th style="text-align: right;">ATS score</th>
<th style="text-align: right;">Keywords</th>
<th style="text-align: right;">Role fit</th>
<th style="text-align: right;">Numbers</th>
<th style="text-align: right;">Recoverable</th>
</tr>
</thead>
<tbody>
<tr>
<td>Data Scientist</td>
<td style="text-align: right;"><strong>52.67</strong></td>
<td style="text-align: right;">27.12</td>
<td style="text-align: right;">14.37</td>
<td style="text-align: right;">0.0</td>
<td style="text-align: right;">+9.72</td>
</tr>
<tr>
<td>Software Engineer</td>
<td style="text-align: right;"><strong>38.42</strong></td>
<td style="text-align: right;">15.33</td>
<td style="text-align: right;">10.04</td>
<td style="text-align: right;">0.59</td>
<td style="text-align: right;">+13.33</td>
</tr>
<tr>
<td>AI / LLM Agent Engineer</td>
<td style="text-align: right;"><strong>34.06</strong></td>
<td style="text-align: right;">12.19</td>
<td style="text-align: right;">9.27</td>
<td style="text-align: right;">0.0</td>
<td style="text-align: right;">+11.66</td>
</tr>
<tr>
<td>Backend / Platform Engineer</td>
<td style="text-align: right;"><strong>32.25</strong></td>
<td style="text-align: right;">10.88</td>
<td style="text-align: right;">9.0</td>
<td style="text-align: right;">0.0</td>
<td style="text-align: right;">+10.88</td>
</tr>
<tr>
<td>Machine Learning Engineer</td>
<td style="text-align: right;"><strong>25.88</strong></td>
<td style="text-align: right;">6.36</td>
<td style="text-align: right;">7.08</td>
<td style="text-align: right;">0.0</td>
<td style="text-align: right;">+12.22</td>
</tr>
</tbody>
</table></div>
<p>Best fit: <strong>data_scientist</strong> (spread across roles: 26.79 points).</p>
<h2 id="resume-ai_llm_agent_engineer-3406100">Resume — ai_llm_agent_engineer (34.06/100)</h2>
<p><strong>AI / LLM Agent Engineer – Agentic LLM Systems with Evaluation and Guardrails</strong></p>
<p>Designed and implemented agentic LLM systems using LangChain ReAct and MongoDB, integrating static and dynamic web scraping, PDF extraction, and tool‑calling workflows. Built AI‑powered natural language search for GitHub repositories, delivering analytics and export capabilities. Developed robust data ingestion pipelines and structured output handling for LLM applications.</p>
<h3 id="projects">Projects</h3>
<ul>
<li>Implemented a LangChain ReAct Agent with Ollama and MongoDB to orchestrate tool‑calling workflows.</li>
<li>Built a static web scraper using requests and BeautifulSoup to provide structured data for LLM inputs.</li>
<li>Developed a dynamic web scraper with Playwright for real‑time data extraction.</li>
<li>Created a PDF extractor converting PDFs to text via pypdf, pymupdf, and OCR for LLM ingestion.</li>
<li>Implemented MongoDB helpers for inserting and searching documents to support LLM retrieval.</li>
<li>Designed an AI‑powered natural language search for GitHub repositories, supporting URL‑based and natural language modes.</li>
<li>Delivered detailed analytics including stats, contributors, language breakdown, commits, and issues for repositories.</li>
<li>Enabled export of analysis results as JSON or CSV for downstream processing.</li>
</ul>
<h3 id="skills">Skills</h3>
<ul>
<li>LangChain</li>
<li>Ollama</li>
<li>MongoDB</li>
<li>requests</li>
<li>BeautifulSoup</li>
<li>Playwright</li>
<li>pypdf</li>
<li>pymupdf</li>
<li>OCR</li>
<li>Docker Compose</li>
<li>Node.js</li>
<li>Express</li>
<li>Multer</li>
</ul>
<h2 id="resume-ml_engineer-2588100">Resume — ml_engineer (25.88/100)</h2>
<p><strong>Machine Learning Engineer – Model Training, Evaluation, and Deployment</strong></p>
<p>Implemented and maintained machine learning pipelines for natural language and web‑scraping tasks, leveraging LangChain, Ollama, and MongoDB. Built data extraction utilities and web scrapers using requests, BeautifulSoup, Playwright, and OCR libraries. Developed backend services for file handling and AI‑powered repository analysis with Docker Compose.</p>
<h3 id="projects_1">Projects</h3>
<ul>
<li>Implemented a LangChain ReAct Agent with Ollama and MongoDB to orchestrate tool calls for AI tasks.</li>
<li>Built a static web scraper using requests and BeautifulSoup for data extraction.</li>
<li>Developed a dynamic web scraper with Playwright for interactive content retrieval.</li>
<li>Created a PDF extractor converting PDFs to text via pypdf, pymupdf, and OCR.</li>
<li>Provided MongoDB helpers for inserting and searching documents in the agent workflow.</li>
<li>Constructed a minimalist file upload system with Node.js and Express, offering web interface, file listing, and download capabilities.</li>
<li>Built a Docker Compose‑based backend and frontend for the local lending application.</li>
<li>Developed an AI‑powered natural language search for GitHub repositories, supporting URL and natural language queries, and delivering analytics in JSON or CSV.</li>
</ul>
<h3 id="skills_1">Skills</h3>
<ul>
<li>LangChain</li>
<li>Ollama</li>
<li>MongoDB</li>
<li>requests</li>
<li>BeautifulSoup</li>
<li>Playwright</li>
<li>pypdf</li>
<li>pymupdf</li>
<li>OCR</li>
<li>Node.js</li>
<li>Express</li>
<li>Docker Compose</li>
</ul>
<h2 id="resume-backend_platform_engineer-3225100">Resume — backend_platform_engineer (32.25/100)</h2>
<p><strong>Backend / Platform Engineer – Service & Infrastructure Builder</strong></p>
<p>Engineered and maintained backend services and infrastructure, focusing on API design, containerization, and data persistence. Delivered microservices with Docker, managed data with MongoDB and PostgreSQL, and implemented robust logging and monitoring solutions.</p>
<h3 id="projects_2">Projects</h3>
<ul>
<li>Implemented a multi-threaded C pipeline for AoA protection, parsing UART frames and validating sensor data using median filtering and outlier rejection.</li>
<li>Designed and executed a finite‑state machine driven by CSV‑defined thresholds to classify flight states and control elevator deflection for AoA mitigation.</li>
<li>Built a minimalist file upload system with Node.js and Express, providing a web interface for uploading, listing, and downloading files, with configurable size limits via Multer.</li>
<li>Developed a dynamic web scraper using Playwright and a static scraper with requests and BeautifulSoup, integrating PDF extraction via pypdf, pymupdf, and OCR.</li>
<li>Created an AI-powered natural language search for GitHub repositories, delivering analytics and exportable results in JSON or CSV formats.</li>
<li>Integrated MongoDB helpers for inserting and searching documents in an agent-based system using LangChain ReAct Agent with Ollama.</li>
<li>Started and built backend and frontend services with Docker Compose, enabling consistent local development environments.</li>
</ul>
<h3 id="skills_2">Skills</h3>
<ul>
<li>C</li>
<li>Node.js</li>
<li>Express</li>
<li>Python</li>
<li>Playwright</li>
<li>BeautifulSoup</li>
<li>requests</li>
<li>pypdf</li>
<li>pymupdf</li>
<li>OCR</li>
<li>MongoDB</li>
<li>Docker Compose</li>
<li>Docker</li>
<li>GitHub</li>
<li>JSON</li>
<li>CSV</li>
</ul>
<h2 id="resume-data_scientist-5267100">Resume — data_scientist (52.67/100)</h2>
<p><strong>Data Scientist – Turning Data into Measured Answers</strong></p>
<p>Built and deployed data pipelines that extract, transform, and analyze information from web sources and code repositories. Delivered AI‑powered search and analytics tools that provide actionable insights in JSON and CSV formats.</p>
<h3 id="projects_3">Projects</h3>
<ul>
<li>Implemented a dynamic web scraper using Playwright to collect structured data for downstream analysis.</li>
<li>Developed a PDF extractor converting documents to text via pypdf, pymupdf, and OCR for data ingestion.</li>
<li>Created MongoDB helper functions to insert and search documents, enabling efficient data retrieval.</li>
<li>Built an AI‑powered natural language search engine for GitHub repositories, delivering detailed analytics such as contributor stats, language breakdown, commits, and issues.</li>
<li>Exported analysis results as JSON or CSV, facilitating integration with downstream data science workflows.</li>
</ul>
<h3 id="skills_3">Skills</h3>
<ul>
<li>Python</li>
<li>Playwright</li>
<li>BeautifulSoup</li>
<li>requests</li>
<li>pypdf</li>
<li>pymupdf</li>
<li>OCR</li>
<li>MongoDB</li>
<li>LangChain</li>
<li>Ollama</li>
<li>Jupyter</li>
<li>pandas</li>
<li>numpy</li>
<li>SQL</li>
<li>data visualization</li>
<li>regression</li>
<li>classification</li>
<li>clustering</li>
<li>hypothesis testing</li>
<li>A/B testing</li>
<li>experiment design</li>
</ul>
<h2 id="resume-balanced-3842100">Resume — balanced (38.42/100)</h2>
<p><strong>Software Engineer – Generalist across services, data, and ML</strong></p>
<p>Implemented multi‑threaded C pipelines for AoA protection, built web‑scraping utilities in Python, and developed a Node.js file upload server. Designed and deployed Docker‑based services, integrated MongoDB helpers, and engineered AI‑powered GitHub repository analytics.</p>
<h3 id="projects_4">Projects</h3>
<ul>
<li>Implemented a multi‑threaded C pipeline for AoA protection</li>
<li>Parsed UART frames and validated sensor data using median filtering and outlier rejection</li>
<li>Estimated AoA with weighted fusion and a Kalman filter</li>
<li>Ran a finite‑state machine driven by CSV‑defined thresholds to classify flight states</li>
<li>Controlled elevator deflection to mitigate excessive AoA in protection mode</li>
<li>Logged state transitions and sensor faults to fault_log.csv</li>
<li>Used LangChain ReAct Agent with Ollama and MongoDB</li>
<li>Provided a static web scraper using requests and BeautifulSoup</li>
<li>Provided a dynamic web scraper using Playwright</li>
<li>Provided a PDF extractor converting PDFs to text via pypdf, pymupdf, and OCR</li>
<li>Included MongoDB helpers for inserting and searching documents</li>
<li>Built a minimalist file upload system with Node.js and Express for FreeBSD</li>
<li>Provided a simple web‑based file upload interface</li>
<li>Stored uploaded files with original filenames</li>
<li>Listed uploaded files with size and upload date</li>
<li>Enabled direct file download from the interface</li>
<li>Configured default maximum upload size of 10 MB via Multer limits</li>
<li>Started and built backend with Docker Compose</li>
<li>Started and built frontend with Docker Compose</li>
<li>Provided AI‑powered natural language search for GitHub repositories</li>
<li>Supported URL‑based and natural language search modes</li>
<li>Delivered detailed analytics including stats, contributors, language breakdown, commits, and issues</li>
<li>Allowed exporting analysis results as JSON or CSV</li>
</ul>
<h3 id="skills_4">Skills</h3>
<ul>
<li>C</li>
<li>Python</li>
<li>Node.js</li>
<li>Express</li>
<li>JavaScript</li>
<li>Docker Compose</li>
<li>MongoDB</li>
<li>LangChain</li>
<li>Ollama</li>
<li>requests</li>
<li>BeautifulSoup</li>
<li>Playwright</li>
<li>pypdf</li>
<li>pymupdf</li>
<li>OCR</li>
<li>Multer</li>
<li>GitHub API</li>
</ul>
<h2 id="improvement-plan">Improvement plan</h2>
<p>5 steps, ~40 hours, +16.9 ATS points if all are completed.</p>
<h3 id="1-add-data-science-notebook-to-aiagentdemo">1. Add Data Science Notebook to AiAgentDemo</h3>
<p><em>12h · +7.23 ATS · role: data_scientist · repo: <code>AiAgentDemo</code></em></p>
<ul>
<li><strong>Gap:</strong> Cannot evidence statistics, visualisation, dataset, notebook, correlation, feature, matplotlib, seaborn, metric, r</li>
<li><strong>Evidence found:</strong> AiAgentDemo (Python web‑scraping and PDF extraction)</li>
<li><strong>Action:</strong> Create a Jupyter notebook (notebook.ipynb) in the AiAgentDemo repository that loads a public dataset (e.g., UCI Wine), performs descriptive statistics, computes correlation matrix, engineers at least two new features, visualises distributions and relationships with matplotlib and seaborn, and records a simple model’s evaluation metric (e.g., accuracy).</li>
<li><strong>Done when:</strong> notebook.ipynb committed to AiAgentDemo with executed cells, plots saved as PNG, and a markdown summary of statistics and metrics.</li>
</ul>
<h3 id="2-create-r-analysis-demo-repository">2. Create R Analysis Demo Repository</h3>
<p><em>10h · +3.95 ATS · role: data_scientist · new work</em></p>
<ul>
<li><strong>Gap:</strong> Missing evidence of R language and related visualisation (ggplot2) for keyword ‘r’</li>
<li><strong>Evidence found:</strong> none</li>
<li><strong>Action:</strong> Initialize a new repository named r-analysis-demo, add an R script (analysis.R) that loads the same dataset, performs summary statistics, computes correlations, creates at least two ggplot2 visualisations, and saves plots as PNG files.</li>
<li><strong>Done when:</strong> analysis.R, summary.txt, and plot PNG files committed to r-analysis-demo.</li>
</ul>
<h3 id="3-update-aiagentdemo-readme-with-data-science-section">3. Update AiAgentDemo README with Data Science Section</h3>
<p><em>4h · +2.63 ATS · role: data_scientist · repo: <code>AiAgentDemo</code></em></p>
<ul>
<li><strong>Gap:</strong> Resume sections lack explicit data‑science artefacts</li>
<li><strong>Evidence found:</strong> AiAgentDemo (existing README with web‑scraping description)</li>
<li><strong>Action:</strong> Edit the README.md to add a “Data Science Projects” section linking to notebook.ipynb, features.py, evaluate_model.py and summarising the statistical and visualisation work.</li>
<li><strong>Done when:</strong> README.md updated and pushed, showing the new section with hyperlinks.</li>
</ul>
<h3 id="4-add-model-metric-reporting-script">4. Add Model Metric Reporting Script</h3>
<p><em>6h · +1.97 ATS · role: data_scientist · repo: <code>AiAgentDemo</code></em></p>
<ul>
<li><strong>Gap:</strong> No demonstrated metric calculation or reporting</li>
<li><strong>Evidence found:</strong> AiAgentDemo (existing Python utilities)</li>
<li><strong>Action:</strong> Create a script (evaluate_model.py) that trains a simple classifier/regressor on the same dataset, computes relevant metrics (e.g., RMSE, F1‑score), and writes a JSON report (metrics.json).</li>
<li><strong>Done when:</strong> evaluate_model.py and generated metrics.json committed to the repo.</li>
</ul>
<h3 id="5-implement-featureengineering-module">5. Implement Feature‑Engineering Module</h3>
<p><em>8h · +1.1 ATS · role: data_scientist · repo: <code>AiAgentDemo</code></em></p>
<ul>
<li><strong>Gap:</strong> Missing evidence of feature creation and statistical processing</li>
<li><strong>Evidence found:</strong> AiAgentDemo (Python code base)</li>
<li><strong>Action:</strong> Add a Python module (features.py) that defines functions for cleaning, scaling, and generating new features from a DataFrame; include unit tests (test_features.py) covering at least three cases.</li>
<li><strong>Done when:</strong> features.py and test_features.py committed, with 100% test pass shown in CI badge or local run output.</li>
</ul>
<h2 id="measure-these">Measure these</h2>
<p>Each of these is plausible but unproven. The agent refused to put a number on it; run the command and the number becomes claimable.</p>
<ul>
<li><code>chronosModelImprovement</code> — The distance‑aware model does not win the majority of series, losing 745 out of 1484.</li>
<li><em>why:</em> contains number(s) absent from the evidence: 745</li>
<li><strong><code>measure this and record the command and result</code></strong></li>
<li><code>tender_search_app</code> — Enables users to add and refine keywords to improve search accuracy.</li>
<li><em>why:</em> asserts an effect without a measured quantity</li>
<li><strong><code>run the evaluation set and record the metric</code></strong></li>
<li><code>sandbox-agent</code> — The agent graph is lazily loaded to improve startup performance.</li>
<li><em>why:</em> asserts an effect without a measured quantity</li>
<li><strong><code>measure this and record the command and result</code></strong></li>
</ul>
<h2 id="refused-claims">Refused claims</h2>
<p>The model proposed these; the evidence did not support them. They are listed so you can see what was filtered, not hidden.</p>
<ul>
<li><code>locallend</code> — The backend is implemented in Java using Spring Boot.<br />
<em>every supporting quote is absent from the evidence</em></li>
<li><code>locallend</code> — The frontend is written in TypeScript with React.<br />
<em>every supporting quote is absent from the evidence</em></li>
<li><code>locallend</code> — The repository includes image upload functionality.<br />
<em>every supporting quote is absent from the evidence</em></li>
<li><code>locallend</code> — The repository implements CRUD operations for items.<br />
<em>every supporting quote is absent from the evidence</em></li>
<li><code>AiTerminal</code> — The project uses the tcell library for terminal UI.<br />
<em>supporting quote is real but does not mention anything in the claim</em></li>
<li><code>AiTerminal</code> — The repository does not include a CI configuration.<br />
<em>supporting quote is real but does not mention anything in the claim</em></li>
</ul>
<h2 id="evidence-index">Evidence index</h2>
<h3 id="aoa_controller_ty">AOA_Controller_ty</h3>
<p>AoA Controller is a multi-threaded C pipeline that protects aircraft from excessive angle‑of‑attack by parsing UART frames, validating sensor data, estimating AoA, and driving elevator commands, warnings, and fault logging.</p>
<ul>
<li>Implements a multi-threaded C pipeline for AoA protection.</li>
<li><code>README.md</code>: "Angle-of-Attack (AoA) protection system for embedded flight control, implemented as a multi-threaded C pipeline."</li>
<li>Parses UART frames and validates sensor data using median filtering and outlier rejection.</li>
<li><code>README.md</code>: "UART_FrameParser │ sem_post(frame_ready_sem) ▼ SensorValidator ← median filter, outlier rejection, confidence score"</li>
<li>Estimates AoA with weighted fusion and a Kalman filter.</li>
<li><code>README.md</code>: "AoA_Estimator ← weighted fusion + Kalman filter"</li>
<li>Runs a finite‑state machine driven by CSV‑defined thresholds to classify flight states.</li>
<li><code>README.md</code>: "FlightControlFSM ← CSV thresholds → NORMAL / WARNING / PROTECTION"</li>
<li>Controls elevator deflection to mitigate excessive AoA in protection mode.</li>
<li><code>README.md</code>: "ElevatorCmdManager ← corrective deflection in PROTECTION mode"</li>
<li>Logs state transitions and sensor faults to fault_log.csv.</li>
<li><code>README.md</code>: "FaultLogger ← logs state transitions + sensor faults"</li>
</ul>
<h3 id="aiagentdemo">AiAgentDemo</h3>
<p>A small demo showing a LangChain ReAct agent that orchestrates static, dynamic, and PDF scraping tools and stores results in MongoDB.</p>
<ul>
<li>Uses LangChain ReAct Agent with Ollama and MongoDB.</li>
<li><code>README.md</code>: "built with <strong>LangChain</strong>, <strong>Ollama</strong>, and <strong>MongoDB</strong>."</li>
<li>Provides a static web scraper using requests and BeautifulSoup.</li>
<li><code>README.md</code>: "static scraper (requests+BS)"</li>
<li>Provides a dynamic web scraper using Playwright.</li>
<li><code>README.md</code>: "dynamic scraper (Playwright)"</li>
<li>Provides a PDF extractor that converts PDFs to text via pypdf, pymupdf, and OCR.</li>
<li><code>README.md</code>: "PDF extractor (pypdf→OCR)"</li>
<li>Includes MongoDB helpers for inserting and searching documents.</li>
<li><code>README.md</code>: "MongoDB tools (save / search)"</li>
<li>The agent only decides which tool to call, delegating work to the underlying functions.</li>
<li><code>README.md</code>: "The agent is a decision layer. It does not scrape, parse, or store data."</li>
</ul>
<h3 id="file-upload-server">file-upload-server</h3>
<p>A minimalist file upload server built with Node.js and Express for FreeBSD.</p>
<ul>
<li>The project is a minimalist file upload system built with Node.js and Express for FreeBSD.</li>
<li><code>README.md</code>: "A minimalist file upload system built with Node.js and Express for FreeBSD systems."</li>
<li>It provides a simple web-based file upload interface.</li>
<li><code>README.md</code>: "Simple web-based file upload interface"</li>
<li>Uploaded files are stored with their original filenames.</li>
<li><code>README.md</code>: "Original filename preservation"</li>
<li>The server lists uploaded files with their size and upload date.</li>
<li><code>README.md</code>: "File listing with size and upload date"</li>
<li>Users can download files directly from the interface.</li>
<li><code>README.md</code>: "Direct file download capability"</li>
<li>The default maximum upload size is 10 MB, configurable via the Multer limits.</li>
<li><code>README.md</code>: "Default maximum file size is 10MB."</li>
</ul>
<h3 id="locallend">locallend</h3>
<p>LocalLend is a full‑stack web application that lets users list, browse, and book items, built with a Java Spring Boot backend and a TypeScript React frontend.</p>
<ul>
<li>The backend is started and built with Docker Compose.</li>
<li><code>README.md</code>: "start and build the backend using docker compose up --build -d (from the locallend directory)"</li>
<li>The frontend is started and built with Docker Compose.</li>
<li><code>README.md</code>: "start and build the frontend using docker compose up --build -d (from the locallend-frontend directory)"</li>
</ul>
<h3 id="githubrepoanalyzer">GithubRepoAnalyzer</h3>
<p>GitHubRepoAnalyzer is a React application that analyzes GitHub repositories using AI-powered natural language search and provides comprehensive analytics.</p>
<ul>
<li>Provides AI‑powered natural language search for GitHub repositories.</li>
<li><code>README.md</code>: "A comprehensive React application for analyzing GitHub repositories with AI-powered natural language search capabilities."</li>
<li>Supports both URL-based and natural language search modes.</li>
<li><code>README.md</code>: "Dual Search Modes - URL Search: Direct repository analysis via GitHub URLs - Natural Language Search: AI-powered repository discovery using plain English descri"</li>
<li>Delivers detailed analytics including stats, contributors, language breakdown, commits, and issues.</li>
<li><code>README.md</code>: "Comprehensive Analytics - Repository statistics and metadata - Contributor analysis with activity metrics - Programming language breakdown with visual charts - "</li>
<li>Allows exporting analysis results as JSON or CSV.</li>
<li><code>README.md</code>: "Export Functionality: Download analysis results in JSON/CSV formats"</li>
<li>Fully containerized with Docker for easy deployment.</li>
<li><code>README.md</code>: "Docker Container (Production Ready)"</li>
<li>Manages GitHub API rate limits with public and private modes.</li>
<li><code>README.md</code>: "Public API Mode: 60 requests per hour (no authentication required) - Private API Mode: 5,000 requests per hour (requires GitHub personal access token)"</li>
</ul>
<h3 id="spring-gemini-chat">spring-gemini-chat</h3>
<p>Spring Gemini Chat is a reactive Spring Boot application that provides real‑time chat powered by Google Gemini AI.</p>
<ul>
<li>Built with Spring WebFlux for high-performance, non-blocking operations</li>
<li><code>README.md</code>: "Built with Spring WebFlux for high-performance, non-blocking operations"</li>
<li>Powered by Google's latest Gemini AI model</li>
<li><code>README.md</code>: "Powered by Google's latest Gemini AI model"</li>
<li>Instant responses with typing indicators</li>
<li><code>README.md</code>: "Instant responses with typing indicators"</li>
<li>Clean, responsive web interface</li>
<li><code>README.md</code>: "Clean, responsive web interface"</li>
<li>Built-in health checks and status indicators</li>
<li><code>README.md</code>: "Built-in health checks and status indicators"</li>
<li>Support for contextual conversations</li>
<li><code>README.md</code>: "Support for contextual conversations"</li>
</ul>
<h3 id="next-express-ejs-crud">next-express-ejs-crud</h3>
<p>A starter that runs a Next.js frontend and an Express + EJS backend side‑by‑side to demonstrate CRUD operations.</p>
<ul>
<li>Provides a backend that serves both EJS‑rendered pages and a JSON API.</li>
<li><code>README.md</code>: "Backend: Express + EJS (server-rendered pages at http://localhost:4000) and a JSON API under /api."</li>
<li>Runs a Next.js frontend that consumes the Express JSON API.</li>
<li><code>README.md</code>: "Frontend: Next.js (at http://localhost:3000) that calls the API."</li>
<li>Rewrites /api/* requests in Next.js to the Express API on port 4000.</li>
<li><code>README.md</code>: "next.config.js rewrites /api/* to http://localhost:4000/api/* in development/production."</li>
<li>Wraps fetch in the frontend to use an absolute URL that works during server‑side rendering.</li>
<li><code>README.md</code>: "web/lib/api.js wraps fetch. On the server it uses an absolute URL based on NEXT_ORIGIN so SSR fetches hit Next first and are rewritten to the API."</li>
<li>Provides a Dockerfile that builds and runs both the Next.js and Express apps in a single container.</li>
<li><code>README.md</code>: "This repo includes a single Dockerfile that builds both apps and runs them in one container (ports 3000 for Next, 4000 for Express)."</li>
<li>Implements a JSON API with CRUD routes for posts.</li>
<li><code>README.md</code>: "API Routes (JSON) - GET /api/posts — list all posts"</li>
</ul>
<h3 id="nextjs-express-crud-hf">nextjs-express-crud-hf</h3>
<p>A full‑stack Next.js and Express CRUD application ready for Hugging Face deployment.</p>
<ul>
<li>Provides a full‑stack CRUD application with server‑side rendering.</li>
<li><code>README.md</code>: "Full-stack CRUD app with SSR, ready for Hugging Face deployment."</li>
<li>Combines Next.js for the front‑end and Express for the API.</li>
<li><code>README.md</code>: "Next.js + Express CRUD Application"</li>
<li>Includes a GitHub Actions workflow that deploys the app to Hugging Face.</li>
<li><code>README.md</code>: "Deploy to Hugging Face: - Add HF_TOKEN to GitHub secrets - Update HF_REPO_ID in workflow - Push to GitHub"</li>
<li>Uses Docker as the SDK for containerized deployment.</li>
<li><code>README.md</code>: "docker"</li>
<li>Configures the application to run on port 7860.</li>
<li><code>README.md</code>: "app_port: 7860"</li>
<li>Runs the front‑end on localhost port 3000.</li>
<li><code>README.md</code>: "Visit http://localhost:3000"</li>
</ul>
<h3 id="aoa_controller_ty_1">AOA_Controller_ty_1</h3>
<p>A real‑time angle of attack monitoring and elevator control system for aircraft, with ESP32 firmware and PC‑based simulation tools.</p>
<ul>
<li>Provides real‑time AOA monitoring and elevator control on ESP32.</li>
<li><code>README.md</code>: "Real-time angle of attack monitoring and elevator control system for aircraft, with ESP32 firmware and PC-based simulation tools."</li>
<li>Simulates all four flight phases using CSV test data.</li>
<li><code>README.md</code>: "Process flight test data from CSV file. Simulates all 4 flight phases (TAKEOFF, CLIMB, CRUISE, LANDING) with realistic FSM state transitions and elevator comman"</li>
<li>Implements a 10°/sec slew‑rate limit on elevator deflection.</li>
<li><code>README.md</code>: "Slew-rate limited elevator (10°/sec max)"</li>
<li>Handles missing sensor readings by treating 0.00 values as absent.</li>
<li><code>README.md</code>: "Handles missing sensor data (0.00 values)"</li>
<li>Provides an interactive UART simulator for real‑time debugging.</li>
<li><code>README.md</code>: "Interactive UART Simulator (simulator.c) ... Real-time interactive testing with manual AOA message input."</li>
<li>Builds without requiring a local ESP‑IDF installation.</li>
<li><code>README.md</code>: "No ESP-IDF dependencies (uses mock headers)"</li>
</ul>
<h3 id="chronosmodelimprovement">chronosModelImprovement</h3>
<p>Chronos Model Improvement implements a distance‑aware enhancement to the Chronos time series forecasting model.</p>
<ul>
<li>Provides a distance‑aware variant of the Chronos forecasting model.</li>
<li><code>README.md</code>: "This repository contains improvements to the Chronos time series forecasting model."</li>
<li>Adds a distance‑aware enhancement to the Chronos model.</li>
<li><code>README.md</code>: "Distance-aware enhancement for Chronos model."</li>
<li>Benchmark shows the distance‑aware model has a higher MAE (1315.20) than the original (1309.50).</li>
<li><code>README.md</code>: "MAE | 1315.20 | 1309.50 | Original"</li>
<li>Benchmark shows the distance‑aware model has a higher RMSE (1593.91) than the original (1584.71).</li>
<li><code>README.md</code>: "RMSE | 1593.91 | 1584.71 | Original"</li>
<li>The distance‑aware model wins 739 out of 1484 series, a win rate of 49.8%.</li>
<li><code>README.md</code>: "Win Rate: Distance-Aware wins 739/1484 series (49.8%)"</li>
</ul>
<h3 id="websocket_mcqs1">websocket_mcqs1</h3>
<p>A MERN‑stack application that lets users create multiple‑choice questions and play against each other with real‑time score updates.</p>
<ul>
<li>The repository implements a real‑time MCQ battle.</li>
<li><code>README.md</code>: "Realtime MCQs battle"</li>
<li>The application is built with the MERN stack.</li>
<li><code>README.md</code>: "It is written in Javascript using MERN stack"</li>
<li>Installation requires running <code>npm install</code>.</li>
<li><code>README.md</code>: "npm install"</li>
<li>The server is started with <code>node server.js</code>.</li>
<li><code>README.md</code>: "node server.js"</li>
<li>Scores are observed in real time by all participants.</li>
<li><code>README.md</code>: "observe their score in realtime"</li>
<li>Users can create MCQs and play against each other.</li>
<li><code>README.md</code>: "create MCQs and play against each other"</li>
</ul>
<h3 id="aniketqwgithubio">aniketqw.github.io</h3>
<p>Aniket Saxena’s personal portfolio site and admin dashboard built with a custom design system.</p>
<ul>
<li>Implemented a custom design system for the portfolio.</li>
<li><code>README.md</code>: "A design system distilled from the personal portfolio of <strong>Aniket Saxena</strong> — AI Engineer & Full-Stack Developer"</li>
<li>Built both a public portfolio site and a private admin dashboard.</li>
<li><code>README.md</code>: "Two surfaces are represented: the public <strong>Portfolio</strong> site and the private <strong>Admin</strong> (repo manager) tool."</li>
<li>Styled the admin dashboard with distinct typography and color accents.</li>
<li><code>README.md</code>: "Admin uses <strong>Title Case with lowercase helper copy</strong> and ALL-CAPS 10–12px uppercase labels with 1–3px letter-spacing."</li>
<li>Implemented a Python script to fetch and filter GitHub repositories.</li>
<li><code>README.md</code>: "Add fetch_repos.py to retrieve and filter GitHub repos"</li>
<li>Added a dynamic projects section that loads data from GitHub.</li>
<li><code>README.md</code>: "Add projects section with dynamic project loading"</li>
<li>Documented design tokens and UI kits for reuse across the site.</li>
<li><code>README.md</code>: "Add design system assets and documentation - colors, typography, components, and UI kits"</li>
</ul>
<h3 id="pehenava-backend">pehenava-backend</h3>
<p>Pehenava Backend is a Node.js/Express application that connects to MongoDB to provide API services for the Pehenava project.</p>
<ul>
<li>The repository implements a Node.js/Express backend that integrates with MongoDB.</li>
<li><code>README.md</code>: "A Node.js/Express backend application with MongoDB integration for the Pehenava project."</li>
<li>It supports a local setup where dependencies are installed via npm.</li>
<li><code>README.md</code>: "When running without Docker, <strong>all dependencies will be installed locally</strong> on your machine."</li>
<li>It also supports a Dockerized deployment that containers all dependencies.</li>
<li><code>README.md</code>: "When running with Docker, <strong>all dependencies are containerized</strong> - nothing is installed on your local machine except Docker itself."</li>
<li>The README provides a Docker command to run a MongoDB container for the backend.</li>
<li><code>README.md</code>: "docker run -d \ --name pehenava-mongo \ -p 27017:27017 \ mongo:latest"</li>
<li>Docker Compose is recommended to manage both the application and MongoDB containers together.</li>
<li><code>README.md</code>: "docker-compose up -d"</li>
<li>The application starts an Express server listening on port 3000.</li>
<li><code>README.md</code>: "Server running on port 3000"</li>
</ul>
<h3 id="aiterminal">AiTerminal</h3>
<p>AiTerminal is a Go-based terminal AI application that runs via <code>go run terminal.go</code>.</p>
<ul>
<li>The application is started by running <code>go run terminal.go</code>.</li>
<li><code>readme.md</code>: "go run terminal.go"</li>
<li>The project uses the gin web framework.</li>
<li><code>go.mod</code>: ""github.com/gin-gonic/gin""</li>
<li>The project uses gorilla/websocket for real‑time communication.</li>
<li><code>go.mod</code>: ""github.com/gorilla/websocket""</li>
<li>The repository does not contain a test suite.</li>
<li><code>signals</code>: ""has_test_suite": false"</li>
</ul>
<h3 id="tender_search_app">tender_search_app</h3>
<p>A web‑based tool that lets users search and filter Indian tenders using a Gradio interface.</p>
<ul>
<li>Provides search functionality across multiple Indian tender sources.</li>
<li><code>README.md</code>: "Search for tenders across multiple sources."</li>
<li>Allows filtering of search results by recency and relevance.</li>
<li><code>README.md</code>: "Filter tenders based on recency and relevance."</li>
<li>Implements a user‑friendly interface using Gradio.</li>
<li><code>README.md</code>: "User-friendly interface built with Gradio."</li>
<li>Builds the application stack with Python, Gradio, Langchain, BeautifulSoup, Requests, and Pandas.</li>
<li><code>README.md</code>: "Technologies Used - Python - Gradio - Langchain - BeautifulSoup - Requests - Pandas"</li>
<li>Runs the application via a Python script that starts the Gradio server.</li>
<li><code>README.md</code>: "Run the application: <code>python main.py</code>"</li>
</ul>
<h3 id="rule-basedcustomkafkapartitioner">Rule-BasedCustomKafkaPartitioner</h3>
<p>A Java 21 demo that implements a custom Kafka partitioner routing messages based on key rules.</p>
<ul>
<li>Implements a custom Kafka partitioner that routes messages based on key rules.</li>
<li><code>README.md</code>: "This project demonstrates a custom Kafka partitioner that routes messages based on rules: <code>USA</code> keys go to the last partition, others are hashed across the rema"</li>
<li>The partitioner sends keys equal to "USA" to the last partition and hashes other keys to the remaining partitions.</li>
<li><code>README.md</code>: "Custom partitioner logic: <code>USA</code> → last partition; others → hash % (numPartitions - 1)."</li>
<li>A producer demo sends sample messages with keys including "USA" to the topic "orders".</li>
<li><code>README.md</code>: "The producer sends sample messages with keys <code>INDIA</code>, <code>GERMANY</code>, <code>USA</code>, <code>FRANCE</code>, <code>USA</code> to topic <code>orders</code>."</li>
<li>Producer output shows "USA" messages are routed to partition 3, the last partition.</li>
<li><code>README.md</code>: "Sent key=INDIA value=Order-1 → partition=1 Sent key=GERMANY value=Order-2 → partition=1 Sent key=FRANCE value=Order-4 → partition=2 Sent key=USA value=Order-3 →"</li>
<li>The demo topic is created with four partitions, enabling the rule‑based routing.</li>
<li><code>README.md</code>: "Topic <code>orders</code> is auto‑created with 4 partitions by default."</li>
<li>Kafka is run locally using Docker Compose for a single‑node setup.</li>
<li><code>README.md</code>: "docker-compose.yml for a single‑node Kafka setup."</li>
</ul>
<h3 id="sandbox-agent">sandbox-agent</h3>
<p>sandbox-agent is a Python-based LangGraph agent framework that supports an educative tool, lazy loading, and web search integration.</p>
<ul>
<li>Credentials for the educative tool are stored in an environment variable.</li>
<li><code>commits/6d8d1e9879</code>: "Move EDUCATIVE_COOKIES_B64 from source code to environment variable"</li>
<li>An educative feature allows users to ask questions about a course using a cookie.</li>
<li><code>commits/f7281212c5</code>: "added educative feature just we can just give cookie and ask question about the course"</li>
<li>A dummy stub container is provided for local tool execution, with a clear error message when run in LangGraph Studio.</li>
<li><code>commits/1a0c192bc9</code>: "added dummy stub container working but some other error NotImplementedError('Tool execution is not available in LangGraph Studio. Please run the agent locally ("</li>
<li>The repository includes a tavily web search tool for external data retrieval.</li>
<li><code>commits/5d6794c559</code>: "updating added tavily web search"</li>
<li>An explicit NotImplementedError informs users that tool execution requires running the agent locally.</li>
<li><code>commits/1a0c192bc9</code>: "NotImplementedError('Tool execution is not available in LangGraph Studio. Please run the agent locally (<code>python harness.py</code>) to execute tools.')"</li>
</ul>
<h3 id="websocket_chatroom">websocket_chatroom</h3>
<p>A simple realtime chatroom implemented with WebSocket.</p>
<ul>
<li>Built a simple chatroom with realtime communication using WebSocket.</li>
<li><code>README.md</code>: "Simple chatroom with realtime communication implemented using websocket"</li>
<li>Implemented broadcasting of messages to multiple clients.</li>
<li><code>commit:3538e568ad</code>: "Broadcast to multiple clients"</li>
<li>Server sends messages back to the client and the client receives them.</li>
<li><code>commit:9fb75c6f01</code>: "Send message back to client from server and receiving it on client side"</li>
<li>Client can send a message to the server via WebSocket.</li>
<li><code>commit:e60c00c599</code>: "send a message from client to server send() (index.html)"</li>
<li>Server responds to the connection event when a client connects.</li>
<li><code>commit:d4c0cf540f</code>: "on connection respond to connection event on server side"</li>
<li>The repository includes an automated test suite.</li>
<li><code>signals</code>: ""has_test_suite": true"</li>
</ul>
<h3 id="chat-starter-code">chat-starter-code</h3>
<p>A simple WebSocket-based chat application built with Node.js.</p>
<ul>
<li>The project is a WebSocket‑based chat application.</li>
<li><code>README.md</code>: "# WebSockets-ws Chat Application"</li>
<li>The application is started with <code>npm install</code> followed by <code>node server.js</code>.</li>
<li><code>README.md</code>: "To run this application, first open up a terminal. Then, run the command below: <code>sh npm install node server.js</code>"</li>
<li>The application is deployed and accessible at https://chat-starter-code.onrender.com/.</li>
<li><code>README.md</code>: "https://chat-starter-code.onrender.com/"</li>
</ul>
<h2 id="run-cost">Run cost</h2>
<ul>
<li>Groq: 11359 tokens across 3 calls</li>
<li>Cache hit rate: 100%</li>
<li>Wall clock: 2779s</li>
</ul>
</section>
<section class="panel report" id="p-trace" role="tabpanel" aria-labelledby="t-trace" hidden>
<a id="trace"></a>
<h1 id="run-trace">Run trace</h1>
<p>A complete record of one pipeline run, written so the output can be <strong>audited rather than trusted</strong>. Every claim the verifier rejected is listed below with its reason — that section is the most informative part of this file.</p>
<ul>
<li><strong>Command:</strong> <code>python scripts/build_sample.py --username aniketqw</code></li>
<li><strong>Account:</strong> <code>aniketqw</code> (public repositories only)</li>
<li><strong>Python:</strong> 3.11.15 on Darwin</li>
<li><strong>Models:</strong> small/mid <code>openai/gpt-oss-20b</code>, big <code>openai/gpt-oss-120b</code></li>
<li><strong>Free-tier ceiling:</strong> 8,000 tokens/min, 1,000 requests/day (per model)</li>
</ul>
<h2 id="stages">Stages</h2>
<div class="scroll"><table>
<thead>
<tr>
<th>Stage</th>
<th>What happened</th>
<th style="text-align: right;">Tokens</th>
<th style="text-align: right;">Seconds</th>
</tr>
</thead>
<tbody>
<tr>
<td>list_repos</td>
<td>99 repositories found via the GitHub MCP server; 57 public. Private repositories are filtered out here, before anything is fetched.</td>
<td style="text-align: right;">—</td>
<td style="text-align: right;">7.7</td>
</tr>
<tr>
<td>select_repos</td>
<td>Jev scored 47 repositories for $0.00161; kept 21. No Groq tokens spent.</td>
<td style="text-align: right;">—</td>
<td style="text-align: right;">50.6</td>
</tr>
<tr>
<td>fetch_facts</td>
<td>Deterministic fetch for 21 repositories (19 with a README, 9 with tests). No model involved.</td>
<td style="text-align: right;">—</td>
<td style="text-align: right;">26.7</td>
</tr>
<tr>
<td>safety_scan</td>
<td>1 finding(s) in the scanned repositories.</td>
<td style="text-align: right;">—</td>
<td style="text-align: right;">0.0</td>
</tr>
<tr>
<td>summarize_repo</td>
<td>21 repositories summarized, 111 raw claims proposed.</td>
<td style="text-align: right;">—</td>
<td style="text-align: right;">0.0</td>
</tr>
<tr>
<td>verify_claims</td>
<td>102/111 claims survived. Pure Python: no model, no tokens.</td>
<td style="text-align: right;">—</td>
<td style="text-align: right;">0.0</td>
</tr>
<tr>
<td>build_resumes</td>
<td>5 resumes written from verified claims only.</td>
<td style="text-align: right;">9180</td>
<td style="text-align: right;">56.0</td>
</tr>
<tr>
<td>plan_improvements</td>
<td>5 evidence-linked steps.</td>
<td style="text-align: right;">2179</td>
<td style="text-align: right;">2637.9</td>
</tr>
</tbody>
</table></div>
<h2 id="headline-numbers">Headline numbers</h2>
<ul>
<li><strong>102/111 claims verified</strong> (92%); 3 need measurement; 6 rejected outright.</li>
<li><strong>5 resumes</strong> built from verified claims only.</li>
<li><strong>Groq:</strong> 11,359 tokens across 3 node(s).</li>
<li><strong>Cache hit rate:</strong> 100% — an unchanged repo costs nothing on a re-run.</li>
<li><strong>Wall clock:</strong> 2779s, most of it waiting on the 8,000 tokens/minute limit.</li>
</ul>
<h2 id="repository-selection">Repository selection</h2>
<p>Scored by Jev (a typed decision model), not by Groq — this costs fractions of a cent and leaves the scarce Groq budget for text.</p>
<div class="scroll"><table>
<thead>
<tr>
<th>Repo</th>
<th style="text-align: right;">Claimable</th>
<th>Tier</th>
<th>Language</th>
</tr>
</thead>
<tbody>
<tr>
<td><code>AOA_Controller_ty</code></td>
<td style="text-align: right;">0.69</td>
<td>mid</td>
<td>C</td>
</tr>
<tr>
<td><code>AiAgentDemo</code></td>
<td style="text-align: right;">0.69</td>
<td>mid</td>
<td>Python</td>
</tr>
<tr>
<td><code>file-upload-server</code></td>
<td style="text-align: right;">0.68</td>
<td>mid</td>
<td>HTML</td>
</tr>
<tr>
<td><code>locallend</code></td>
<td style="text-align: right;">0.67</td>
<td>mid</td>
<td>Java</td>
</tr>
<tr>
<td><code>GithubRepoAnalyzer</code></td>
<td style="text-align: right;">0.64</td>
<td>mid</td>
<td>JavaScript</td>
</tr>
<tr>
<td><code>spring-gemini-chat</code></td>
<td style="text-align: right;">0.64</td>
<td>mid</td>
<td>HTML</td>
</tr>
<tr>
<td><code>next-express-ejs-crud</code></td>
<td style="text-align: right;">0.63</td>
<td>mid</td>
<td>JavaScript</td>
</tr>
<tr>
<td><code>nextjs-express-crud-hf</code></td>
<td style="text-align: right;">0.60</td>
<td>mid</td>
<td>JavaScript</td>
</tr>
<tr>
<td><code>AOA_Controller_ty_1</code></td>
<td style="text-align: right;">0.59</td>
<td>mid</td>
<td>C</td>
</tr>
<tr>
<td><code>chronosModelImprovement</code></td>
<td style="text-align: right;">0.59</td>
<td>mid</td>
<td>Python</td>
</tr>
<tr>
<td><code>websocket_mcqs1</code></td>
<td style="text-align: right;">0.58</td>
<td>small</td>
<td>JavaScript</td>
</tr>
<tr>
<td><code>aniketqw.github.io</code></td>
<td style="text-align: right;">0.57</td>
<td>mid</td>
<td>HTML</td>
</tr>
<tr>
<td><code>pehenava-backend</code></td>
<td style="text-align: right;">0.57</td>
<td>mid</td>
<td>TypeScript</td>
</tr>
<tr>
<td><code>AiTerminal</code></td>
<td style="text-align: right;">0.56</td>
<td>small</td>
<td>Go</td>
</tr>
<tr>
<td><code>tender_search_app</code></td>
<td style="text-align: right;">0.53</td>
<td>small</td>
<td>Python</td>
</tr>
<tr>
<td><code>Rule-BasedCustomKafkaPartitioner</code></td>
<td style="text-align: right;">0.52</td>
<td>small</td>
<td>Java</td>
</tr>
<tr>
<td><code>Booking-System</code></td>
<td style="text-align: right;">0.50</td>
<td>mid</td>
<td>Python</td>
</tr>
<tr>
<td><code>sandbox-agent</code></td>
<td style="text-align: right;">0.50</td>
<td>mid</td>
<td>Python</td>
</tr>
<tr>
<td><code>TaskAssignWithDjango</code></td>
<td style="text-align: right;">0.48</td>
<td>big</td>
<td>Python</td>
</tr>
<tr>
<td><code>websocket_chatroom</code></td>
<td style="text-align: right;">0.48</td>
<td>big</td>
<td>HTML</td>
</tr>
<tr>
<td><code>chat-starter-code</code></td>
<td style="text-align: right;">0.46</td>
<td>small</td>
<td>HTML</td>
</tr>
</tbody>
</table></div>
<h2 id="claim-verdicts">Claim verdicts</h2>
<p>Verification is <strong>plain Python</strong>, not a prompt. Every number in a claim must appear in that repository's evidence bundle; every quote must be present <em>and</em> relevant to the claim it supports.</p>
<h3 id="rejected-6">Rejected (6)</h3>
<p>These were proposed by the model and thrown out. They never reach a resume.</p>
<ul>
<li><code>locallend</code> — The backend is implemented in Java using Spring Boot.</li>
<li><em>every supporting quote is absent from the evidence</em></li>
<li><code>locallend</code> — The frontend is written in TypeScript with React.</li>
<li><em>every supporting quote is absent from the evidence</em></li>
<li><code>locallend</code> — The repository includes image upload functionality.</li>
<li><em>every supporting quote is absent from the evidence</em></li>
<li><code>locallend</code> — The repository implements CRUD operations for items.</li>
<li><em>every supporting quote is absent from the evidence</em></li>
<li><code>AiTerminal</code> — The project uses the tcell library for terminal UI.</li>
<li><em>supporting quote is real but does not mention anything in the claim</em></li>
<li><code>AiTerminal</code> — The repository does not include a CI configuration.</li>
<li><em>supporting quote is real but does not mention anything in the claim</em></li>
</ul>
<h3 id="needs-measurement-3">Needs measurement (3)</h3>
<p>Plausible but unproven. The agent refused to attach a number it could not find, and emitted a measurement task instead.</p>
<ul>
<li><code>chronosModelImprovement</code> — The distance‑aware model does not win the majority of series, losing 745 out of 1484.</li>
<li><em>contains number(s) absent from the evidence: 745</em> → <strong><code>measure this and record the command and result</code></strong></li>
<li><code>tender_search_app</code> — Enables users to add and refine keywords to improve search accuracy.</li>
<li><em>asserts an effect without a measured quantity</em> → <strong><code>run the evaluation set and record the metric</code></strong></li>
<li><code>sandbox-agent</code> — The agent graph is lazily loaded to improve startup performance.</li>
<li><em>asserts an effect without a measured quantity</em> → <strong><code>measure this and record the command and result</code></strong></li>
</ul>
<h2 id="resume-scores">Resume scores</h2>
<p>Deterministic 100-point ATS rubric, no model involved.</p>
<div class="scroll"><table>
<thead>
<tr>
<th>Role</th>
<th style="text-align: right;">Score</th>
<th style="text-align: right;">Keywords</th>
<th style="text-align: right;">Role fit</th>
<th style="text-align: right;">Numbers</th>
</tr>
</thead>
<tbody>
<tr>
<td>data_scientist</td>
<td style="text-align: right;"><strong>52.67</strong></td>
<td style="text-align: right;">27.12</td>
<td style="text-align: right;">14.37</td>
<td style="text-align: right;">0.0</td>
</tr>
<tr>
<td>balanced</td>
<td style="text-align: right;"><strong>38.42</strong></td>
<td style="text-align: right;">15.33</td>
<td style="text-align: right;">10.04</td>
<td style="text-align: right;">0.59</td>
</tr>
<tr>
<td>ai_llm_agent_engineer</td>
<td style="text-align: right;"><strong>34.06</strong></td>
<td style="text-align: right;">12.19</td>
<td style="text-align: right;">9.27</td>
<td style="text-align: right;">0.0</td>
</tr>
<tr>
<td>backend_platform_engineer</td>
<td style="text-align: right;"><strong>32.25</strong></td>
<td style="text-align: right;">10.88</td>
<td style="text-align: right;">9.0</td>
<td style="text-align: right;">0.0</td>
</tr>
<tr>
<td>ml_engineer</td>
<td style="text-align: right;"><strong>25.88</strong></td>
<td style="text-align: right;">6.36</td>
<td style="text-align: right;">7.08</td>
<td style="text-align: right;">0.0</td>
</tr>
</tbody>
</table></div>
<h3 id="bullets-dropped-during-resume-writing">Bullets dropped during resume writing</h3>
<p>None. Every generated bullet traced back to a verified claim.</p>
<h2 id="improvement-plan">Improvement plan</h2>
<p>5 steps, ~40 hours, +16.9 ATS points. The gain figure comes from the same rubric that produced the score, not from the model.</p>
<div class="scroll"><table>
<thead>
<tr>
<th>Step</th>
<th>Repo</th>
<th style="text-align: right;">Hours</th>
<th style="text-align: right;">ATS gain</th>
</tr>
</thead>
<tbody>
<tr>
<td>Add Data Science Notebook to AiAgentDemo</td>
<td><code>AiAgentDemo</code></td>
<td style="text-align: right;">12</td>
<td style="text-align: right;">+7.23</td>
</tr>
<tr>
<td>Create R Analysis Demo Repository</td>
<td><code>new work</code></td>
<td style="text-align: right;">10</td>
<td style="text-align: right;">+3.95</td>
</tr>
<tr>
<td>Update AiAgentDemo README with Data Science Section</td>
<td><code>AiAgentDemo</code></td>
<td style="text-align: right;">4</td>
<td style="text-align: right;">+2.63</td>
</tr>
<tr>
<td>Add Model Metric Reporting Script</td>
<td><code>AiAgentDemo</code></td>
<td style="text-align: right;">6</td>
<td style="text-align: right;">+1.97</td>
</tr>
<tr>
<td>Implement Feature‑Engineering Module</td>
<td><code>AiAgentDemo</code></td>
<td style="text-align: right;">8</td>
<td style="text-align: right;">+1.1</td>
</tr>
</tbody>
</table></div>
<h2 id="security-findings-in-the-scanned-repositories">Security findings in the scanned repositories</h2>
<ul>
<li><strong>[high]</strong> <code>pehenava-backend/pehnava-ai-frontend/.env.local</code> — An .env file is committed. These normally hold live credentials.</li>
</ul>
<h2 id="measured-quality">Measured quality</h2>
<div class="scroll"><table>
<thead>
<tr>
<th>Metric</th>
<th style="text-align: right;">Result</th>
</tr>
</thead>
<tbody>
<tr>
<td>claim faithfulness (text that reaches a resume)</td>
<td style="text-align: right;">n/a</td>
</tr>
<tr>
<td>number precision</td>
<td style="text-align: right;">1.0</td>
</tr>
<tr>
<td>whole-summary faithfulness (includes unverified prose)</td>
<td style="text-align: right;">0.9184</td>
</tr>
<tr>
<td>repositories judged</td>
<td style="text-align: right;">6</td>
</tr>
</tbody>
</table></div>
<h2 id="notes">Notes</h2>
<ul>
<li>Private repositories are excluded before any fetch, so no private content enters this pipeline at all.</li>
<li>Summaries are cached by (repo, head SHA, prompt version, model); the cache hit rate above shows how much of this run was free.</li>
</ul>
<h2 id="how-to-reproduce">How to reproduce</h2>
<pre><code class="language-bash">pip install -e ".[dev]"
cp .env.sample .env # add GROQ_API_KEY, TYPESAFE_API_KEY, GITHUB_PERSONAL_ACCESS_TOKEN
rba doctor
python scripts/build_sample.py --username aniketqw
</code></pre>
<p>Summaries are cached by <code>(repo, head SHA, prompt version, model)</code>, so a second run over unchanged repositories spends no tokens at all.</p>
</section>
<section class="panel report" id="p-how" role="tabpanel" aria-labelledby="t-how" hidden>
<h2 id="how-it-works">How it works</h2>
<pre><code>list_repos official GitHub MCP server, read-only toolset
select_repos Jev scores every repo for "is this claimable"; no Groq tokens
fetch_facts deterministic: README, tree, commits, tests, CI, dependencies
safety_scan flags .env / tokens / cookies committed in the repositories
summarize_repo strict JSON, one repo per call, rate limited
verify_claims CODE, not a prompt: unsupported claims are dropped
score_roles deterministic 100-point ATS rubric
build_resumes bullets drawn only from verified claims
plan_improvements evidence-linked steps with hours and ATS gain
</code></pre>
<h3 id="why-claims-survive-or-die">Why claims survive or die</h3>
<p><code>verify_claims</code> is plain Python. Every digit in a generated claim must appear
in that repository's evidence bundle, and every supporting quote must be both
present <strong>and</strong> relevant to the claim it backs. A claim that fails is not
softened — it is dropped, and reappears as a measurement task.</p>
<p>The clearest example from a real run: the model wrote <em>"losing 745 out of
1484"</em>, having computed <code>1484 &minus; 739</code> itself. The evidence contains 739 and
1484 but never 745, so the claim was refused.</p>
<h3 id="measured-quality">Measured quality</h3>
<p>Claim faithfulness — the text that actually reaches a resume — and
number precision both measure <strong>1.00</strong>. Whole-summary faithfulness is <strong>0.83</strong>,
and that gap is reported rather than averaged away: the descriptive
<code>problem</code> and <code>approach</code> fields are framing, are not run through
<code>verify_claims</code>, and can contain fair-sounding inference.</p>
<h3 id="reproducibility-measured">Reproducibility, measured</h3>
<p>Two independent runs over the same repositories, both at <code>temperature: 0</code>:</p>
<div class="scroll"><table>
<thead>
<tr>
<th>Half of the pipeline</th>
<th>Result</th>
</tr>
</thead>
<tbody>
<tr>
<td>Verification (<code>verify_claims</code>)</td>
<td><strong>Identical.</strong> Same 102/111 verdicts, the same 6 rejections, the same reasons.</td>
</tr>
<tr>
<td>Generation (<code>build_resumes</code>)</td>
<td><strong>Variable.</strong> ATS scores moved by up to <strong>24 points</strong> (<code>balanced</code> 59.1 → 34.8).</td>
</tr>
</tbody>
</table></div>
<p>That split is the design working as intended, not an accident: the half that
decides what may be <em>said</em> is deterministic Python, so it cannot drift. The
half that decides <em>how</em> to say it is a language model, and Groq's <code>gpt-oss</code>
models are not reproducible even at temperature 0.</p>
<p>The practical consequence is worth stating plainly: <strong>treat a single ATS score
as indicative, not precise.</strong> What is stable is which claims are allowed
through — and that is the property the project actually promises.</p>
<h3 id="operating-inside-a-free-tier">Operating inside a free tier</h3>
<p>The binding constraint is <strong>8,000 tokens per minute</strong>, not the request rate.
One evidence-linked repository summary costs about 2,074 tokens, so the ceiling
is roughly four per minute. Responses are cached by
<code>(repo, head SHA, prompt version, model)</code>, so an unchanged repository costs
nothing on a re-run.</p>
</section>
<footer>
Built by <a href="https://github.com/aniketqw/resume-builder-agent">resume-builder-agent</a>
· LangGraph + Groq + the official GitHub MCP server
· page generated by <code>scripts/build_site.py</code>
</footer>
</div>
<script>
const tabs = [...document.querySelectorAll('[role=tab]')];
tabs.forEach(tab => tab.addEventListener('click', () => {
tabs.forEach(other => {
const selected = other === tab;
other.setAttribute('aria-selected', selected);
document.getElementById(other.getAttribute('aria-controls')).hidden = !selected;
});
}));
</script>
</body>
</html>
|