|
Download README.md from DataDrivenAlex/python-genai-portfolio: direct link, hf CLI and curl.
- Browser
- Download file 3.72 kB
-
https://huggingface.co/spaces/DataDrivenAlex/python-genai-portfolio/resolve/main/README.md
- Command line
-
hf download hf://spaces/DataDrivenAlex/python-genai-portfolio/README.md
-
curl -L -o README.md https://huggingface.co/spaces/DataDrivenAlex/python-genai-portfolio/resolve/main/README.md
3.72 kB
| title: Alexandru Cojocari — Python Backend & GenAI Portfolio | |
| emoji: 🧩 | |
| colorFrom: blue | |
| colorTo: purple | |
| sdk: static | |
| app_file: index.html | |
| # Alexandru Cojocari | |
| **Python Backend Engineer · GenAI & Applied AI · QA Automation** | |
| I have around 10 years of professional Python backend experience and commercial Java experience. My portfolio explores reliable backend systems and practical AI engineering, with QA automation as a secondary strength. | |
| I build small, runnable demos around the engineering details that matter: safe interfaces, reproducible evaluation, data integrity, and useful tests. These are portfolio projects, not claims of production deployment. | |
| [GitHub profile](https://github.com/DatadrivenAlex) | |
| ## Selected projects | |
| ### AI and GenAI engineering | |
| - [Model Rollout Gate](https://github.com/DatadrivenAlex/model-rollout-gate) — Offline champion–challenger checks for binary models: paired bootstrap intervals, segment gates, an empirical p95 latency budget, and a replayable JSON decision. Uses synthetic data; it informs a human release decision and does not deploy models. | |
| - [LLM Gateway Workbench](https://github.com/DatadrivenAlex/llm-gateway-workbench) — FastAPI demo of request idempotency, per-tenant quotas, provider adapters, bounded retries and fallback. The included provider is deterministic/offline; this is not a production gateway. | |
| - [Document AI Extractor](https://github.com/DatadrivenAlex/document-ai-extractor) — Extracts invoice and receipt fields from PDFs/images with pdfplumber, Tesseract OCR, rules, optional provider fallback, and Pydantic financial checks. Scoring uses a synthetic corpus; the optional live API is not exercised in CI. | |
| - [AI Security Guardrails](https://github.com/DatadrivenAlex/ai-security-guardrails) — Rule-based HTTP firewall that can allow, deny, or rewrite text before or after an LLM call. Offline and deterministic; it makes no model calls and is not a production WAF. | |
| - [NL2SQL Analytics Agent](https://github.com/DatadrivenAlex/nl2sql-analytics-agent) — Read-only SQL analytics over a small SQLite demo warehouse, with SQLGlot parsing and SQLite authorizer safeguards. The default path uses a fixed offline question catalog; an OpenAI-compatible endpoint is optional. Not a production text-to-SQL system. | |
| - [Semantic Search & Embeddings](https://github.com/DatadrivenAlex/semantic-search-embeddings) — Compares BM25, FAISS dense retrieval, and reciprocal-rank fusion on a project-authored FAQ corpus. The checked-in evaluation uses a deterministic hashing embedder; CI does not run the optional neural model. | |
| ### Backend, data and quality | |
| - [ML Dataset Lineage Registry](https://github.com/DatadrivenAlex/ml-dataset-lineage-registry) — FastAPI service for schema-validated, immutable dataset snapshots with canonical SHA-256 fingerprints and parent lineage. Uses SQLite and synthetic data; it has no authentication or tenant isolation. | |
| - [Playwright QA Automation](https://github.com/DatadrivenAlex/playwright-qa-automation) — Self-contained Python test framework for a demo shop: httpx API tests, Playwright UI tests, page objects, pytest, and failure artifacts. It does not contact an external store or perform payments. | |
| ## Engineering focus | |
| Python · FastAPI · Java · SQL · API design · GenAI integration · retrieval and evaluation · data validation · model rollout checks · security guardrails · pytest · Playwright · CI | |
| ## About the demos | |
| The repositories are intended for learning, review, and portfolio discussion. Each README documents its own assumptions and limits. Unless a project says otherwise, examples use synthetic or local demo data; do not send confidential data to these demos. | |