Akanksha User

At Garmin, I design and ship production AI agents for a diagnostics platform using LangGraph and TypeScript. I moved an agent from prototype to a core product feature in under three months and built full-stack automation that reduced manual diagnostic time by 60%.

I build agents with multi-step reasoning, tool use, self-correction, evals, regression testing, and golden datasets. My production agent work has reached 92% task completion accuracy while hybrid agent and rule-based designs reduced API costs by 35%.

Previously at HCLTech, I built AI-powered automation agents processing 40K–80K transactions per day, reducing manual intervention by 45%. I also designed Azure event-driven microservices, decomposed a legacy monolith into 12 independently deployable services, and improved payment-system reliability through idempotency, retries, dead-letter queues, Saga patterns, and Transactional Outbox.

I enjoy forward-deployed engineering: translating real diagnostic and workflow problems into reliable, measurable customer solutions.