AI Systems Engineer building reliable agent infrastructure, RAG pipelines, and practical GenAI products.
I build AI applications that connect models, tools, data, and automation into dependable workflows. My focus is agent reliability, MCP integrations, model routing, RAG/data synchronization, observability, security boundaries, and privacy-first AI.
- DeltaVec: incremental PostgreSQL-to-Qdrant synchronization with change detection, deterministic chunk IDs, orphan cleanup, tombstones, and reconciliation sweeps.
- Customer Support Agent: FastAPI + LangGraph workflow with a deterministic policy gate that the LLM cannot override.
- Anvit: privacy-first, offline document assistant for PDF and Word files on-device.
- WebNexus: self-contained crawling, document search, embeddings, and retrieval workflow.
Python, FastAPI, PostgreSQL, Qdrant, Docker, Dagster, Linux, LangGraph, MCP, vector search, embeddings, and cloud deployment tooling.
I start with failure modes such as stale data, duplicate work, unsafe tool calls, provider outages, unclear permissions, and misleading health checks. Then I make the system observable and testable before adding more automation.
I share practical notes about agent reliability, RAG/data synchronization, self-hosting, and building AI products that solve real problems.
AI engineering roles, focused freelance projects, and collaborations involving agent reliability, MCP/tool integrations, RAG/data systems, or practical automation.