Work

Software Engineer

Evals Lead

Joined after the Groq–NVIDIA partnership. Worked on inference and evaluation systems at the intersection of hardware and model performance.

Head of Evals

Led the Evals team. Built openbench, an open-source standard for running evals easily, reliably, and in a reproducible manner. Designed evaluation infrastructure that standardized benchmarking across 20+ evaluation suites and became the backbone of Groq's model quality process.

Researcher in Residence

Developed synthetic data pipelines for training language models. Built systems that generated high-quality training data at scale, contributing to Nous's open-source model releases.

Provider-agnostic, open-source evaluation infrastructure for language models. Standardized benchmarking across 20+ evaluation suites.

An evaluation framework using debate simulations to assess AI models' reasoning and communication skills.

A multimodal benchmark for testing vision capabilities and reasoning in AI models.