I’m a Founding Engineer, AI Platform based in New York. I build the systems behind AI products: agents, retrieval, realtime voice, evaluation, and the platform that keeps them running.
My job is to get AI past the impressive demo: sound architecture, honest evaluation, clear interfaces, and fewer surprises in production.
Coffee status: probably reheating it.
Production and open source
The work behind HistOracle’s historical intelligence and AI platform, followed by a few public tools you can inspect.
I built and now lead the intelligence layer that makes historical figures answer with evidence, stay faithful to their period, and hold up under adversarial evaluation. The retrieval and evaluation strategy is built in-house and runs on the production path. In a controlled pilot, it reached 94.1% factual accuracy.
I built and maintain the platform behind HistOracle: one operating layer for tenant isolation, model and voice-provider routing, access control, usage economics, observability, and recovery. It is the foundation we use to ship and support historical AI experiences for more than 2,000 users.
A macOS app for private voice dictation and meeting notes. Transcription, speaker diarization, and summaries run locally, so conversations stay off the cloud.
A personal knowledge base maintained by an LLM. It compiles sources once at ingest, builds a durable wiki, and answers later questions from that stored knowledge.
An agent skill for editing the stock phrases, inflated claims, and repetitive rhythms out of AI-written prose. It turns a public style guide into a practical checklist that coding agents can follow.
Older experiments and smaller tools are still available on GitHub.
Explore all repositoriesWhere I’ve worked
HistOracle · Edison, NJ
I lead the build and day-to-day operation of HistOracle’s AI platform, from historical intelligence and evaluation to realtime voice, tenant isolation, and production infrastructure.
Built historical intelligence and evaluation systems, agent workflows, and voice experiences that became core parts of the production platform.
Stevens Institute of Technology · Hoboken, NJ
Built and evaluated more than 11 agent systems, automated academic workflows, and helped teams improve orchestration and retrieval.
BlueTree Consultancy Services · Bengaluru, India
Built data automation and real-time analytics for enterprise teams, with API governance and zero-downtime releases.
How I work
I like the messy part between a capable model and a product people can rely on.
I work across agent orchestration, retrieval, streaming, evaluation, and the interface around the model. Users should know what the system did and why.
I pay close attention to privacy, latency, observability, and failure modes. Those details decide whether an AI feature survives contact with real users.
Agentic systems · RAG · evaluation · prompt architecture · model routing
Python · TypeScript · Rust · FastAPI · PostgreSQL · event-driven services
GCP · AWS · Azure · containers · CI/CD · observability · local-first AI
Prototyping · system design · UX · technical leadership · shipping
Writing
Short pieces on the parts of AI engineering that become interesting after the demo works.
I’d rather publish when I have something useful to say. The first pieces will cover RAG evaluation, realtime voice, and local AI.
Let’s build something useful