Spaces:
Runtime error
Runtime error
File size: 5,653 Bytes
b91193c | 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 | #!/usr/bin/env python3
"""Populate Redrob idea submission PPTX from current team content."""
from __future__ import annotations
import argparse
import shutil
from pathlib import Path
from pptx import Presentation
from pptx.util import Pt
PROJECT_ROOT = Path(__file__).resolve().parent.parent
DEFAULT_TEMPLATE = PROJECT_ROOT / "challenge" / "Idea Submission Template _ Redrob.pptx"
def _set_text(shape, text: str, *, font_pt: int = 14) -> None:
tf = shape.text_frame
tf.clear()
p = tf.paragraphs[0]
p.text = text
p.font.size = Pt(font_pt)
def build_deck(template: Path, output: Path) -> None:
prs = Presentation(str(template))
# Slide 1 β title fields (text boxes 1β3)
s1 = prs.slides[0].shapes
_set_text(s1[1], "Team Name : Sarva Automata", font_pt=20)
_set_text(
s1[2],
"Problem Statement : Rank 100K candidates for Senior AI Engineer "
"(search/retrieval/ranking); NDCG@10 focus; CPU-only; honeypot-aware.",
font_pt=16,
)
_set_text(s1[3], "Team Leader Name : Arjun Mangarath", font_pt=18)
slide_bodies = [
# Slide 2
"Hybrid recall (BM25 + BGE dense + career RRF) β LightGBM LambdaRank β "
"BGE cross-encoder on top-700 with tuned fusion.\n\n"
"Differentiation: IR + learned ranking + Redrob behavioral signalsβnot keywords. "
"Explicit trap/honeypot and availability handling.",
# Slide 3
"JD: 5β8 YOE production ML, search/retrieval/ranking, embeddings, India hubs, "
"product over research, recruiter reachability.\n\n"
"Signals: silver tiers, LTR features, career evidence, GitHub, trap flags, "
"behavioral twins (response, notice, recency). Cross-encoder for semantic fitβno runtime LLM.",
# Slide 4
"Retrieve ~2K (BM25 + FAISS + career RRF) β LTR + penalties β CE rerank top-700 β "
"fusion + modifiers β top-100 CSV.\n\n"
"Models: BM25, FAISS, LightGBM, BGE-reranker-base, Optuna-tuned fusion/modifiers.\n\n"
"Combine: weighted CE/LTR/RRF Γ behavioral multiplier β trap/clone/research penalties.",
# Slide 5
"Reasoning: 5β6 templates from whitelisted fields (YOE, skills, location, availability, "
"production vs research).\n\n"
"No LLM at rank time β no hallucinated justifications.\n\n"
"Traps: hard-exclude honeypots; clone cap top-30; availability down-rank; research penalty.",
# Slide 6
"Offline: preprocess β features/embeddings/BM25/FAISS β silver labels β train LTR β "
"tune modifiers (~90 min).\n\n"
"Online: rank.py loads artifacts β recall β LTR β CE β fusion/modifiers β "
"team_sarva_automata.csv + audit.",
# Slide 7
"Offline: preprocess.py, train_ltr.py, tune_modifiers.py\n"
"Online: rank.py + HF Docker sandbox\n"
"Artifacts: parquet features, FAISS, BM25, ltr_model.lgb, fusion_params, modifier_params",
# Slide 8
"Audit: proxy NDCG@10 = 1.0; NDCG@30 = 0.9922; tier-5 top 10/30/100 = 9/26/70; "
"0 honeypots in top-100; 0 low-availability in top-30.\n\n"
"Constraints: precomputed indexes; rank ~5β8 min CPU, 16GB RAM, no network at inference.",
# Slide 9
"Python, Polars, LightGBM, sentence-transformers, FAISS, rank-bm25, FastAPI, Optuna. "
"Gemini offline (JD + labels). Cursor for engineering assist.",
# Slide 10
"GitHub: github.com/ashokbugude/recruiter-candidate\n"
"Sandbox: huggingface.co/spaces/ashokbugude/redrob-ranker "
"(upload JSONL/JSON β€100 β ranked CSV)\n"
"Portal CSV: team_sarva_automata.csv\n"
"Reproduce: python rank.py --candidates ./challenge/candidates.jsonl "
"--out ./team_sarva_automata.csv",
]
for slide_idx, body in enumerate(slide_bodies, start=1):
shapes = prs.slides[slide_idx].shapes
content_shape = shapes[-1]
_set_text(content_shape, body, font_pt=13)
output.parent.mkdir(parents=True, exist_ok=True)
prs.save(str(output))
def main() -> int:
parser = argparse.ArgumentParser(description="Build methodology PPTX for portal upload.")
parser.add_argument("--template", type=Path, default=DEFAULT_TEMPLATE)
parser.add_argument(
"--out",
type=Path,
default=DEFAULT_TEMPLATE,
help="Output .pptx (default: overwrite template in challenge/)",
)
parser.add_argument(
"--export-pdf",
type=Path,
default=None,
help="Optional PDF path via PowerPoint COM (Windows only)",
)
args = parser.parse_args()
if args.out.resolve() != args.template.resolve():
shutil.copy2(args.template, args.out)
build_deck(args.template if args.out == args.template else args.out, args.out)
print(f"Wrote deck: {args.out}")
if args.export_pdf:
try:
import comtypes.client # type: ignore
pdf_path = args.export_pdf.resolve()
pdf_path.parent.mkdir(parents=True, exist_ok=True)
powerpoint = comtypes.client.CreateObject("Powerpoint.Application")
powerpoint.Visible = 1
deck = powerpoint.Presentations.Open(str(args.out.resolve()), WithWindow=False)
deck.SaveAs(str(pdf_path), 32)
deck.Close()
powerpoint.Quit()
print(f"Wrote PDF: {pdf_path}")
except Exception as exc:
print(f"PDF export skipped ({exc}). Export manually from PowerPoint: Save as PDF")
return 0
if __name__ == "__main__":
raise SystemExit(main())
|