Relayer Founder 2025 – present

I believe the interfaces we use to interact with agents are the bottleneck for broader AGI adoption, and in the future the scope of agents and swarms will be magnitudes larger than humans can comprehend through a chat message.

- Built an agent as an interface in a phone app, released the alpha and got positive feedback, but continual cloud agents were too expensive at the time.

- Built an agent harness that replaces chat with a graph, built on top of recursive language models (relayerlabs.ai/relayer-agent).

Ontos Technical Advisor

I helped a pair of exciting founders go 0 to 1 on their new startup, including engineering, agent development and hiring. Ontos sells project analysis for the pre-construction process using agent swarms.

Microsoft AI

I was a lead across the three engineering pillars behind Ads Engineering at Microsoft AI: Ranking, Index and Data. I used this unique cross-system expertise to ship improvements to the algorithm and system at scale.

- Went 0 to 1 on building brand showroom agents. We trained a custom model to generate the shopping pages and built a C++ inference service to run the agents in under a second.

- Created the LLM-generated ads infrastructure for Bing and Copilot ads, from retrieval to policy adherence and generation. This powers Performance Max Ads.

AI Research 2017 – 2019

- Built a recreation of the Raven II surgical robot and environment for performing surgery in virtual reality, as an RL environment for training autonomous surgery models.

- Added facial recognition and basic information retrieval to an augmented reality application prototype for health care assistants.