Designing, building, and delivering high-fidelity reinforcement learning environments used by leading AI labs to train models on real-world software engineering tasks, owning the full lifecycle from concept and architecture to deployment and iteration while driving product and technical strategy.

KnockoutAI is an intelligent candidate screening service that keeps hiring managers in the loop, ensuring excellent hiring outcomes while letting you hire 10x as fast. Better than traditional ATS systems, automatically screen and categorize thousands of candidates at once.

Reduced Azure Databricks costs by 50%, implemented Postgres optimizations that led to 10x read/write improvements for customer-facing applications, led development and adoption of agentic AI tools in the energy management space.

Data engineer building multi-lingual data collection pipelines to train speech models for automated drive-thru experiences at Watson Orders. Led the development of an ensemble speech recognition model, which leverages LoRA-tuned SOTA speech-to-text models to generate high-quality annotations used to train downstream ML models.

Designed training, inference and eval of highly configurable, cross-domain LLM agents.

Data engineer building multi-lingual data collection pipelines to train machine learning models for automated drive-thru experiences. Deployed language identification RNN-T model to identify useful Spanish audio from thousands of hours of daily audio in production. This data was labelled, enriched, and structured in a way that enabled the development of bilingual order-taking models in McDonald's drive-thrus.