At Morningstar, I build production LLM systems for the Intelligence Engine Platform, including agentic workflows that orchestrate GPT and Claude across three production AI features. I also fine-tuned and quantized small language models, achieving ~80% lower inference cost at comparable accuracy.
I built document ingestion pipelines that extract structured data, tables, and charts from PDFs, reducing manual data-preparation effort by ~35%. In my ModelForge project, I built a control plane for deploying models and Docker-based apps to users’ AWS accounts, with cost-aware GPU sizing and scheduled start and stop.
