Member of Technical Staff at OpenAI

Architected a production-grade LLM inference platform on Kubernetes, improving scalability by 40% while maintaining 99.9%+ availability. Built GPU-aware scheduling, intelligent request routing, asynchronous execution, and tuned throughput and latency using techniques such as continuous batching, distributed caching, streaming inference, and KV-cache optimization. Productized reusable capabilities for RAG, AI agents, MCP, structured tool/function calling, and foundation-model integrations to cut AI application delivery time by 50% across 15+ engineering teams. Operationalized deployments with automated evaluation/benchmarking/validation and secure serving APIs. Implemented observability with OpenTelemetry, Prometheus, Grafana, tracing, and centralized logging, reducing troubleshooting time by 45%. Improved resilience with canary releases, health checks, fault isolation, capacity controls, and self-healing recovery, sustaining 99.9%+ availability.

Staff Software Engineer at Stripe

Modernized mission-critical payment infrastructure using Go, Java, Kubernetes, Kafka, PostgreSQL, Redis, and cloud-native services, increasing scalability by ~35%. Decomposed transaction workflows into independently deployable services and asynchronous pipelines, increasing processing capacity by ~28%. Defined engineering standards for service contracts, API governance, infrastructure-as-code, deployment automation, and observability, reducing recurring engineering effort by ~25%. Built safeguards for capacity management, health validation, graceful degradation, fault isolation, and recovery, reducing operational incidents by ~30%. Coordinated CI/CD, automated testing, and release controls to shorten deployment cycles by ~30%, and guided architecture decisions balancing performance, cost, security, availability, and maintainability to improve productivity by ~20%.

Senior Software Engineer at Stripe

Engineered high-throughput transaction systems in Go using Kafka, PostgreSQL, and Redis, handling millions of secure daily transactions at 99.99% availability. Reworked transaction processing around concurrency, message ordering, idempotency, retries, and failure recovery, increasing peak throughput by ~30%. Tuned PostgreSQL access via query-plan analysis, indexing, schema optimization, connection pooling, and targeted Redis caching, cutting critical API latency by ~27%. Profiled compute/database/messaging/cache workloads to identify bottlenecks, increasing backend efficiency by ~25%. Traced production failures using metrics, application tracing, health telemetry, and distributed diagnostics, reducing recurring downtime incidents by ~35%.

Software Engineer at Stripe

Delivered Java and Go backend applications and REST APIs for payment authorization, transaction processing, and internal service integrations, increasing application capacity by ~25%. Created reusable service libraries and API components for common payment capabilities, reducing implementation effort for new features by ~20%. Refactored PostgreSQL schemas/indexes and SQL/persistence layers for transaction-heavy workloads, reducing database response times by ~30%. Expanded unit and integration testing and CI, reducing regression-related production issues by ~25%. Refactored legacy backend components into modular services with clear interfaces and standardized practices, accelerating feature delivery by ~20%.