Lead SDE (ML) — Persistent Systems
I build systems that read documents the way a specialist would — fine-tuning LLMs and VLMs, designing multi-modal retrieval pipelines, and extracting structured knowledge graphs from messy medical, legal, and financial data. Most of my work lives at the intersection of small-data training and large-scale document understanding, turning unstructured records into something a machine can actually query.
Python C++ SQL openCypher
PyTorch Transformers spaCy LangGraph vLLM
Qdrant FAISS Neo4j
Docker Vertex AI SageMaker Bedrock
Building pipelines that turn unstructured chaos into structured knowledge.