I build ML-powered products and the systems that make them dependable—from document ingestion and hybrid retrieval to local LLM inference, evaluation, APIs, and deployment.
- 🔎 Working on RAG, dense + sparse retrieval, grounded generation, and agentic tool use
- ⚙️ Comfortable across the stack: Python, Go, FastAPI, PostgreSQL, Qdrant, Docker, and Kubernetes
- 🧪 I care about reproducibility, evaluation, observability, and tests, not only impressive demos
- 📐 Going deeper into probability, optimization, linear algebra, representation learning, and statistical ML
- 📚 Current research interests: retrieval quality, efficient inference, evaluation, and reliable ML systems
End-to-end system for document Q&A with citations and cross-document contradiction detection. Highlights: weighted dense + sparse retrieval, Qdrant RRF fusion, background indexing, configurable local/cloud inference.
| Personal-finance platform with grounded recommendations, voice chat, and a planner → tools → finalizer pipeline. Highlights: allow-listed RAG + web search, STT/TTS, user-scoped data tools, 150+ integration tests.
|
Deployable support automation stack designed for local GPU inference and flexible hybrid infrastructure. Highlights: vLLM, TEI embeddings, Qdrant, MinIO, PostgreSQL, and Graylog behind one deployment interface.
| Research-oriented cognitive training and measurement engine with deterministic experimental protocols. Highlights: trial-level data, millisecond timing, reproducible seeds, raw CSV/JSON export, analytics-first design.
|
| Modeling & retrieval | ML systems | Data & infrastructure |
|---|---|---|
| PyTorch · TensorFlow · embeddings | FastAPI · vLLM · TEI | PostgreSQL · Qdrant · MinIO |
| RAG · hybrid search · reranking | model serving · async workers | Docker · Kubernetes · Terraform |
| evaluation · grounded generation | APIs · observability · testing | Linux · CI/CD · cloud/hybrid deploys |
Interested in ML Engineer, Applied ML, LLM/RAG Engineer, and ML Platform roles.
If you are building useful, rigorous AI systems, I would love to connect.
