ML Engineer & Researcher · M.S. CS @ ASU (Tempe)
I work on retrieval systems that actually scale — RAG pipelines over million-table corpora, cross-modal reasoning, and the infrastructure plumbing that keeps it from falling over in production. Previously built migration tooling and observability infra at Oracle.
CRAFT — Cascaded Retrieval for Tabular QA
Training-free cascaded retrieval architecture for large-scale tabular question answering.
- 🟡 33× embedding cost reduction
- 🏆 SOTA on large-scale benchmarks
Production RAG · CoRAL Lab
FAISS-HNSW indexing + hybrid sparse/dense retrieval over 1M+ tables. The boring parts — latency, cost, quantized inference — matter too.
- 🟢 Scaled retrieval to 1Million+ tables,passages while maintaining 87%+ Recall@10 and 96%+ Recall@50
| Project | What it is |
|---|---|
| Edge Face Recognition | MTCNN + FaceNet on AWS Greengrass · INT8 quantization · 3.7× speedup |
| True RNG | Entropy harvested from ambient audio. Genuinely fun to think about. |
| Stick Hero Bot | OpenCV automation. First CV project. Still proud of it. |
| Polymarket | WIP — prediction market data exploration. |
Open to discussing retrieval systems, reproducing CRAFT experiments, or just talking about why dense retrieval alone isn't enough.