Omkar Satve

I’m a Machine Learning Engineer with 3.5 years of experience in cloud-native ML architecture, NLP/LLM model development, and scalable ML/GenAI service delivery. I focus on end-to-end ownership—from building reliable inference systems to maintaining production-grade ML infrastructure.

At Morningstar, I led productionization and deployment of an LLM-powered service with a 6-step prompt pipeline, QA gates, and observability, serving millions of companies in PitchBook’s ACRS pipeline. I re-architected the ML inference layer by replacing eager initialization of 150+ models with lazy loading and shifting to process-based parallelism, reducing median latency from 25s to 4s and P99 from 120s to 20s.

I also took ownership and production support of Fund Resolution and Industry Service (ACRS), ensuring continuity and operational readiness. By validating Cronjob schedules, monitoring dashboards, API health, and error handling—and running load tests with Locust—I helped ensure reliable system behavior across environments.

Previously at Exela Technologies, I developed and deployed an automated healthcare contract extraction system using LLMs, reducing manual processing time by 60%. I’ve also built RAG chatbot experiences (using Qdrant and LLM providers) and optimized a RoBERTa-based customer support ticket classifier, improving decision-making time by 36%, while supporting production integration for faster workflows.