Sriram Krishna

About

Backend/AI engineer who has built production LLM orchestration and agentic workflow systems in Python/FastAPI on Kubernetes across AWS/Azure. Demonstrated strong reliability engineering by debugging a real-world memory retention issue that caused latency spikes/timeouts, and strong data/performance chops with a PostgreSQL optimization that cut query latency from ~1.2s to ~15ms. Targets roles building scalable, guardrailed AI-driven workflow automation with robust observability and human-in-the-loop controls.

Key Strengths

• Built and operated production LLM orchestration/conversation service (hundreds to low-thousands RPM, bursty traffic)
• Diagnosed and fixed memory retention causing latency spikes/timeouts via heap inspection and reference analysis
• Refactored async/background-task patterns to control object lifetimes and reduce memory churn
• End-to-end ownership of LLM workflow assistant backend (tool calling, schemas, guardrails, validation, latency optimization)
• Database performance tuning with measurable impact (slow query reduced ~1.2s to ~15ms using indexing + query rewrite)
• Strong production engineering practices: layered architecture, typing, testing pyramid, and incident debugging with logs/metrics/tracing
• Designed AI production guardrails (drift monitoring, fallbacks, human-in-loop, audit logs) including moderation edge-case handling

Experience

Education

Skills

Languages