AI Agent Engineer · RAG Systems · LLM Harness · Inference Optimization · NVIDIA / Edge AI
I work on the engineering layer behind AI products — the part that makes LLM systems reliable, measurable, and useful in real workflows.
My main areas are:
- AI agents and tool-using workflows
- RAG systems over documents, APIs, and business data
- MCP-based agent tooling and integrations
- NVIDIA / Jetson inference and edge AI workflows
- LLM evaluation, reliability, and observability
- Backend infrastructure for AI applications
Most AI projects do not fail because the model is weak. They fail because the system around the model is slow, unreliable, hard to evaluate, or difficult to connect with real workflows.
That is where I focus.
| Project | What it does |
|---|---|
| financial-agent | AI financial research assistant with a modern TypeScript application structure. |
| agent-class | AI agent examples covering LangChain, RAG agents, task agents, LangGraph, SQL agents, n8n workflows, local LLM tool-calling, and evaluation patterns. |
| ChatRTX-MCP-Integration | Local RAG and MCP integration work around NVIDIA ChatRTX, TensorRT-LLM, LlamaIndex, FAISS, voice input, and document querying. |
| ai-agent-roadmap | Structured AI agent reference covering agent design, evaluation, security, deployment, and MCP concepts. |
| Project | What it does |
|---|---|
| Jetson-conveyor | Jetson Orin Nano computer vision system using YOLO, CUDA, ONNX, OpenCV, and custom detection logic for conveyor-based foil detection. |
| Project | What it does |
|---|---|
| pdf2doc | Document conversion utility for PDF-to-document workflows. |
| receipt-OCR | OCR-based receipt processing experiment. |
| WebAutomation | Browser and task automation work. |
Systems over prompts.
Reliability over hype.
Evaluation over guessing.
Production behavior over demo behavior.
A useful AI agent should retrieve the right context, call the right tools, handle failure cases, and leave enough traces to understand what happened.
If an AI system cannot be tested, measured, or improved, it is not ready for production.
- production-grade AI agents
- RAG systems over documents and business data
- MCP-based tool integrations
- NVIDIA / Jetson inference workflows
- backend infrastructure for AI products
- turning AI prototypes into real working systems
