ML Engineer Practitioner building RAG systems, agentic workflows, and end-to-end ML pipelines.
Georgia Tech OMSCS (ML specialization) · 2 years production SWE at JPMorgan · Bilingual EN/ES
Ask my TwinBot about my projects, skills, or experience.
🔍 Modular RAG
Production-oriented RAG framework. Hybrid retrieval (BM25 + FAISS + RRF + MMR), explicit wiring, observability-first. Computation and control separated by design.
🍽️ Plate2Recipe
Multimodal pipeline: ViT ingredient recognition → GPT-2/LSTM recipe generation.
Key finding: lower training loss (100k samples) produced worse outputs than a smaller, better-tuned run (10k samples) — quality ≠ loss.
📋 HireSignal
Schema-first job posting extraction pipeline — two LLM calls, everything else deterministic Python.
Pydantic-enforced structured extraction, Jinja2 rendering, QA audit gate, alias-aware skill match scoring, and a SQLite job tracker. Handles employment, freelance, and internship postings. Streamlit UI + CLI.
🔭 Currently working on: adding retrieval evaluation (Hits@k, MRR) to Modular-RAG
💬 Ask me about: RAG system design · hybrid retrieval · reliable agentic pipelines
📫 Reach me: cordova.nellie@outlook.com.com · LinkedIn
🌐 Portfolio: cordovank.github.io
📄 Resume: ML Engineer
🔍 Open to: ML Engineer · AI Engineer · LLM Engineer roles