Kerim T.

I'm a Senior AI Engineer who builds agentic workflows and voice AI that connect language models to real tools and data, including in healthcare. At Curo.ai, I wrote the entire platform for a HIPAA-compliant medical voice agent that books patient appointments by phone. It handled about 1,000 after-hours calls a week, and that success led to a phase 2 planned at 10,000-15,000 calls a week. At Databricks, I built HIPAA-compliant Agent Bricks and Genie workspaces that connect natural-language questions to tools that query the data, adopted by a 35-person team. For a national security customer, I built a hybrid search platform that powers LLMs, designing an ontology first and then building a knowledge graph and vector store from its documents. I also migrated a major media and entertainment company's AI/ML platform, improving model AUC from about 0.62 to 0.69. Before that, at Riot Games, I architected a petabyte-scale Databricks lakehouse handling 30M+ daily events and cut annual compute spend by $3M. Today I run my own self-hosted AI platform, where local coding agents execute typed, verified plans. I'm also a co-author on a vLLM fix that's in review, and I reported a gpt-oss bug in OpenHands that has a fix in review.