Building Autonomous Multi-Agent Systems for the Enterprise.
I engineer resilient GenAI workflows using LangGraph, MCP, and Agentic RAG. Moving beyond simple chatbots to create observable, evaluated, cost-aware, and self-correcting AI architectures.
I engineer resilient GenAI workflows using LangGraph, MCP, and Agentic RAG. Moving beyond simple chatbots to create observable, evaluated, cost-aware, and self-correcting AI architectures.
Technical Proficiency
Selected projects demonstrating architectural patterns for GenAI.
A Hierarchical Multi-Agent System that decomposes vague requests into executable actions. Features a top-level Supervisor that orchestrates specialized Sub-Agents (SQL, Calendar, Mail) to execute complex workflows without context pollution.
An autonomous web-research agent built on the Model Context Protocol (MCP). It separates the LLM client from the tool execution server, communicating via SSE (Server-Sent Events).
Beyond linear RAG: An Agentic Retrieval System that reasons before searching. Features ephemeral vector storage for strict data privacy and session isolation.
Deploying Multimodal Agent
My next milestone: An AI Auditor capable of watching and listening to video recordings. It leverages Gemini 2.5 to process long-context multimedia and uses Pydantic to enforce Structured JSON Output for automated compliance checks and sentiment analysis.
I leverage 10+ years of Systems Engineering experience to build resilient, production-grade GenAI architectures.
My focus moves beyond simple chatbots to the infrastructure around the model:
observability (custom middleware),
evaluation (eval-gates on retrieval & faithfulness),
unit economics (granular FinOps), and scalability (MCP standards).
This portfolio demonstrates the architectural rigor and attention to detail that
define my engineering approach.