Building systems where LLMs stop being demos and start being infrastructure.
Senior AI Engineer at BlackLine (FinTech), previously at Amazon where I built 0→1 automated patching pipelines and ML-based early warning systems for AWS service reliability. Master's in AI from University of Cincinnati. AWS Certified ML Engineer.
📍 San Francisco Bay Area · 🚀 Open to AI Engineering roles
Multi-agent research assistant built at the MongoDB Agentic Orchestration Hackathon (Cerebral Valley, SF). Three specialized agents search grants, papers, and news sources in parallel — coordinated by a LangGraph orchestrator with deterministic ranking and MongoDB Atlas for intelligent result caching. OpenAI handles keyword extraction; the system degrades gracefully when individual agents fail.
LangGraph · MongoDB Atlas · OpenAI · Python
Production-deployed forecasting assistant on Streamlit Community Cloud. Accepts arbitrary time series data and returns forecasts with natural-language explanations — built for analysts who shouldn't need to know what SARIMA is.
LangGraph · Multi-Agent System · RAG · Python · LLMs · Time Series
Conversational business intelligence layer over structured product data. RAG pipeline with ChromaDB embeddings, LangGraph for multi-turn memory, GPT-4o-mini for generation. Designed to answer questions a sales rep would actually ask, not just "summarize this table."
LangGraph · Agentic Workflow · Advanced RAG · Critical Thinking · Conversation Streaming · Advanced Chatting Features · AI Explainability · ChromaDB · Multi-model support · RAG Evaluations · Human Feedback and Confidence Scoring · Python
DataScienceGPT is an AI copilot that compresses that entire Data Exploration and ML solution development into minutes. AI-powered Data Science team that can create ML solutions by applying Data Science techniques and employing a bunch of Data Science agents to do the job.
LangGraph (multi-agent orchestration) · RAG with ChromaDB · FLAML AutoML · OpenAI · Tool Calling · Plotly · Model Explainability · Streamlit · Multi-model support · Human-in-the-loop · Python
- Hakadorio — Contributing to a local-first task manager with AI-powered task management (Python/FastAPI backend + TypeScript frontend). Built with Anthropic's Claude API for natural language task creation and management.
- ResoPrism — Co-created at the MongoDB Agentic Orchestration Hackathon (Cerebral Valley, SF). Multi-agent research assistant using LangGraph orchestration, OpenAI-powered keyword extraction, and MongoDB Atlas for intelligent caching. Agents search grants, papers, and news in parallel with deterministic ranking.
Job Application Agent — Fully automated application pipeline using LangGraph, n8n, and OpenClaw on a self-hosted Mac Mini. Reads job postings, tailors resume/cover letter via Claude API, submits — with a human-in-the-loop review step before anything goes out.
Local AI Infrastructure — n8n + Ollama + Claude API running on personal hardware. Exploring what changes when inference has no rate limits, no latency overhead, and no per-token cost.
Content Automation — LLM pipelines that turn rough notes into polished LinkedIn posts, maintaining voice consistency across drafts.
New Updates every month.
- Long-horizon agentic task execution and failure recovery strategies
- Structured output reliability in production RAG pipelines
- Evaluation frameworks for multi-agent systems (beyond vibes-based testing)
- Foundation model fine-tuning for domain-specific reasoning
- Context caching, advanced RAG techniques to make RAG systems more powerful.
AI / ML
Languages & Frameworks
Cloud & Data
|
|
|
|
