Samuel Cook

I’m a Senior Software Engineer with 6 years of applied machine learning and production ML systems expertise, focused on model distillation, reinforcement learning, and large-scale multimodal data mining. I architect teacher-student pipelines and deploy compressed encoders to optimize inference latency and memory footprint on AWS and GCP.

At Nexstrat, I reduced model size by 8x and inference latency by 72% across a 200-node GPU cluster. I built a high-throughput pipeline indexing petabytes of video and lidar with BigQuery, PostgreSQL, and Redis—enabling sub-second queries over 95M assets—and I scaled serving to support 10k QPS with 95th percentile latency of 18ms.

I’ve also shipped end-to-end systems at SAP and JPMorgan Chase & Co: multimodal feature encoders that improved retrieval recall by 31%, automated evaluation frameworks across 200+ distillation experiments, and backend/ETL services that reduced data availability latency for retraining from 24 hours to 2 hours. I bring an engineering-first approach—CI/CD, observability, A/B testing, and mentoring—so models perform reliably in production.