I design and deploy end-to-end intelligent systems: LLM and RAG pipelines, agentic AI, MLOps, and computer vision. I take products from architecture to production on AWS and GCP using Python, LangChain, LangGraph, FastAPI, and Docker.

• I architect backend AI systems for data ingestion, processing, and retrieval across enterprise use cases.
• I build RAG pipelines with a focus on contextual retrieval, prompt engineering, and response grounding.
• I deploy production infrastructure on AWS and GCP with CI/CD, containerization, and observability.
• I work across the stack: LLMs (OpenAI, Anthropic Claude), agentic orchestration, vector search, and computer vision.

Most of my recent work lives in private company repositories, so the highlights below describe what I have built rather than link to source.

Cairn — Open Source Agentic incident-analysis copilot: an agent orchestrator, a cost-and-sensitivity-aware model router, MCP tool servers for observability and runbooks, and an approval-gated write path, backed by a Postgres+pgvector data layer and a scenario-based evaluation harness. Ships with a companion npm CLI that installs into any project for platform analysis, a live local dashboard, and chatting with your logs.
Stack: Python, Node.js, MCP, PostgreSQL/pgvector, Terraform, Kubernetes, OPA
Source · npm

FedBid Intelligence Platform AI-powered federal contracting intelligence platform on AWS. Orchestrated LangGraph agents with OpenAI and Anthropic Claude to automate opportunity discovery, document parsing (PDF, DOCX, XLSX), compliance checking, and multi-factor bid/no-bid scoring across 10,000+ government contract opportunities.

• Built a semantic relevance matching engine using OpenAI text-embedding-3-small (1536-dim) with pgvector on PostgreSQL, scoring opportunity-to-company fit with a weighted formula across capability, past performance, competitiveness, and size fit.
• Engineered an event-driven ingestion pipeline (AWS Step Functions, Lambda, SQS, ECS Fargate) pulling federal data from SAM.gov, FPDS, USAspending, and GSA CALC+ on a 4-hour refresh cycle with idempotent upserts and S3-backed storage.
• Shipped MLOps pipelines with GitHub Actions CI/CD, Docker multi-platform builds to AWS ECR, and automated ECS Fargate deployments; a centralized agent registry enables independent deployment of 7+ AI agents.
• Ran production infrastructure (ECS Fargate, ALB, RDS PostgreSQL, S3, WAF, Route 53) with auto-scaling, RBAC, async DB access, and Langfuse tracing, keeping LLM costs under $165 per month.
  Stack: Python, LangGraph, OpenAI, Anthropic Claude, pgvector, PostgreSQL, AWS, Docker, Langfuse

Enterprise Insight Engine High-performance platform for enterprise data search and contextual AI. Integrated data from Snowflake and Populi, using MCP agents and LLM-based prompt expansion with MongoDB schema awareness to sharpen query intent and retrieval quality. Optimized context filtering for token efficiency in high-throughput search scenarios.
Stack: Python, LLMs, MongoDB, Snowflake, MCP

Agenda Architect FastAPI backend integrated with MongoDB and Qdrant for persistence and semantic vector search. Deployed LLMs locally via NVIDIA NIM to cut latency and cost versus cloud inference, with a LangChain RAG pipeline that generates structured multi-section agendas from user input. Orchestrated backend, database, vector store, and local LLM services with Docker Compose.
Stack: FastAPI, LangChain, Qdrant, MongoDB, NVIDIA NIM, Docker

CreatorMatch Engine Built an ETL pipeline on GCP to ingest and transform complex JSON data into BigQuery. Developed a Health Score algorithm from engagement and credibility metrics, plus a recommender engine to support marketing sponsorship decisions.
Stack: GCP, BigQuery, Python

RoadVision GeoLocator Deep learning system to geolocate US roadway assets from 2D images and vehicle-mounted footage. Fused Vision Transformer features with metadata and depth maps, improving geolocation accuracy by over 50 percent, from 17ft error to under 7ft, through model fine-tuning and depth estimation.
Stack: PyTorch, Vision Transformers, OpenCV

Languages

AI and Machine Learning

Data and Vector Stores

MLOps and Cloud

Web