Jon Ukmata

At ISA Consulting / myNIME, I researched and prototyped an AI notification summary feature, turning raw notification data into structured prompt inputs for 30-day LLM-generated summaries.

I implemented the summary flow across a Next.js frontend, backend APIs, and an AI service integration, with caching and fallback behavior. I also created a Langfuse prompt evaluation workflow using prompt variants and LLM-as-a-Judge criteria.

I supported insurance document extraction evaluation by preparing ground-truth outputs, reviewing AI responses, and suggesting prompt improvements. I also contributed to Notifications V2 backend work, including notification types, threaded grouping, database relationships, and SQL migrations.

For my EstateIQ university project, I built backend, database, authentication, and task-planning features, then evaluated house-price prediction with Linear Regression and XGBoost. I’m interested in production-ready AI, LLM applications, data engineering, and enterprise AI systems.