At Perplexity, I developed infrastructure for LLM experimentation, post-training, automated evaluation, and inference workflows. I built distributed Python and PyTorch pipelines for model experimentation, data processing, and compute-intensive ML workloads.
At OpenAI, I developed model training and evaluation infrastructure and systems for measuring model capabilities, quality, regressions, reliability, and behavior across model versions. I collaborated with research and infrastructure engineers to productionize experimental ML techniques and improve throughput, reproducibility, observability, and infrastructure efficiency.
Earlier, at Datadog, I developed components of a cloud observability and distributed monitoring platform, including Go and Python services, Kafka processing pipelines, and React/TypeScript interfaces. At Microsoft, I developed cloud-based enterprise applications and contributed to API design, database optimization, and cloud deployment pipelines.
