Capital One
Software Engineer Intern
- Cut LLM inference queue wait by 35% across 1M+ daily requests by architecting a distributed scheduling service with priority-based queuing.
- Built a multi-threaded orchestration layer in Fargate that processes inference requests concurrently, raising GPU utilization by 60%.
- Shipped an OpenTelemetry and New Relic dashboard used by 750+ analysts and engineers to monitor GPU efficiency, reducing incident diagnosis time by 15%.
