Selected work

Agents we putinto production.

Not demos. Systems running real traffic, with evals, observability, and on-call.

Northwind Copilot
Internal agent

Northwind Copilot

A retrieval-grounded ops copilot for a logistics team — answers from 40k SOP docs.

RAG · Qdrant · Next.js
Vela Eval Harness
LLM quality

Vela Eval Harness

Continuous eval + regression gating for a fintech assistant before every deploy.

Braintrust · CI/CD
Atlas Voice Agent
Voice · realtime

Atlas Voice Agent

A low-latency voice agent for inbound support, sub-300ms turn-taking.

Realtime · LiveKit
Meridian Pipeline
Data + RAG

Meridian Pipeline

Document ingestion and reranking pipeline powering a legal research agent.

Python · reranking
Forge Deploy
ML platform

Forge Deploy

Production deploy, observability, and rollback for a multi-agent system.

K8s · Datadog