AI & infrastructure engineer. I build the machinery around models that makes them reliable in production — from schema to ship to the on-call rotation.
Currently at Surface Labs (YC'23). Previously Amazon and UCSF × Stanford.
Multi-model orchestration across 4 frontier models with validation, retry, and structured output — built for production reliability, not demos.
Content idea → AI orchestration → Validate + recover → Structured output
A high-reliability pipeline on Kafka, Redis, and Cassandra — circuit-breakers and idempotent writes for clinical-grade systems where a duplicate event is a patient-safety risk.
Kafka → Redis cache → Cassandra → FastAPI
Event-driven validation and recovery on S3 and DynamoDB for 2M+ records — catching bad data at the door and eliminating 70% of manual toil across the ingest path.
S3 event → Validate pipeline → DynamoDB → Reconcile + CI/CD
AI phone agents that take orders and answer questions, integrating ElevenLabs, Clover POS, Twilio, and Stripe.
Twilio call → ElevenLabs agent + RAG → Clover POS → Supabase + Stripe
I'm Amisha, an AI and infrastructure engineer studying Computer Science and Data Science at UC Berkeley. I build the systems that have to keep working once they become load-bearing.
My range runs from RAG and retrieval systems at Surface Labs to high-reliability healthcare pipelines with UCSF and Stanford, data recovery at Amazon, and founding-engineer work on voice agents at Cognova. The throughline is reliability: idempotent writes, retry-aware queues, and evaluation that runs on every deploy.
I care about the seam where technical depth meets product judgment: taking an idea from schema to ship to the on-call rotation, and leaving the system simpler than I found it.
Let's build something worth trusting.
Open to founding & product-engineering roles — from schema to ship to the on-call rotation.