Open source

I contribute to the open-source infrastructure AI products run on: agent frameworks, RAG engines, LLM serving, and the Apache Arrow and DataFusion data stack.

Upstream projects: 12
Stars across those projects: 250k+
Downloads a month: 340M+
Languages: Go, Python, Rust

Why I work upstream

Many AI products run on the same open-source layers.

Agent frameworks, serving layers and query engines sit under a lot of AI products. I read their internals and send changes upstream where they fall short. A merged change reaches every team that uses the project.

My writing and talks cover the same ground, from agent harnesses to evals.

Read the writing See the talks

Recognition

Work a project called out by name.

Aug 2026Agno

Community MVP, August 2026

Agno is one of the most widely used open-source frameworks for building agent systems. Teams build agents with it, serve them through its runtime, and keep the data and the control plane on their own infrastructure.

GitHub stars: 42k+
Downloads a month: 2M+
Forks: 5.8k+

I contributed to the Cohere integration in Agno. Agno named it in the August community roundup (opens in new tab).

Agent frameworksLLM providers

Read the announcement View the pull request

Where I contribute

Projects with my changes merged upstream, by the layer of the AI stack they serve.

Agents and RAG

Where agents call tools and pull in context. A bug here changes what the model sees, or what it is allowed to do.

LLM serving

The layer that loads models, caches attention state and translates between provider APIs.

Data systems

The columnar format and query engine that analytics and feature pipelines read through.

Evals and tooling

Where teams measure agent quality and build the code around it.

Stars and downloads rounded down, as of October 2026. Downloads count each project's main package on PyPI, crates.io or npm over the last month.

Activity

Public commits, pull requests, reviews and issues on GitHub over the last year.

842 contributions on GitHub in the last year

Follow on GitHub