Software Engineer at DoorDash

Architected an AWS + Kubernetes (EKS) platform for production LLM and vision workloads, reducing inference latency via GPU resource optimization. Built reusable Terraform modules and GitHub Actions CI/CD pipelines to automate provisioning across multi-account AWS environments, reducing deployment time by 30%. Optimized model serving using PyTorch, vLLM, and Triton by tuning GPU scheduling and autoscaling, improving utilization by 30% and lowering compute costs by 18%. Implemented observability with Datadog/Prometheus/CloudWatch for GPU performance and service health, reducing troubleshooting time by 35%. Designed centralized LLM/Agent gateways and OPA-based guardrails for consistent security and governance. Collaborated cross-functionally to deploy AI services with AWS IAM and IaC, maintaining 99.9% availability across multiple production releases.

Software Engineer at Capital One

Integrated FastAPI-based backend services and REST APIs, connecting PostgreSQL and Amazon DynamoDB to support internal applications and improve API response times by ~20%. Assisted in RAG pipeline development to improve search relevance over 10K+ enterprise records. Optimized frequently used SQL queries and added Redis caching, reducing endpoint latency by ~25%. Supported deployments and CI/CD using Docker, AWS, and Jenkins, contributing to ~15% faster release cycles. Worked in an Agile team (10+ members) on code reviews, debugging, and third-party API integrations across 5+ sprint releases.