Projects
CTO and cofounder
The enterprise agent platform.
Companies connect their documents, chats, and email to Coresix. Their teams then get sourced answers and agents that work on that data. The data stays under the company’s control: every query is logged, and customer data is never used to train models. As CTO, I lead engineering and built most of the platform: the multi-tenant product, the retrieval layer over company data, the agent runtime and its guardrails, and the security controls for enterprise deployments.
Open source
Author
An agent workflow that measures how much you can trust it.
A workflow in which AI coding agents build GitHub issues, run by a person or an AI orchestrator. A builder agent makes the change, fixed checks run, and a verifier agent that cannot edit code links each acceptance criterion to evidence. On seeded test PRs, the verifier caught 5 of 5 bad PRs and passed 3 of 3 good ones, at about $0.40 a run. In a pilot on ten real issues, 9 PRs merged with no human edits, for $8.43 in total.
Curator
A curated list of benchmarks, harnesses, grading methods, and safety evals for AI agents that use tools and act over many steps.
Curator
A curated list of RAG frameworks, vector stores, document processing, security and privacy, and evaluation tools.