Emre Turan
AI Engineer — agent safety & human-in-the-loop systems
6+ years building LLM agents, multi-agent systems, and production AI. B.S. Statistics & Data Science, UCLA.
Now
Headroom — teams gate their AI agents with human-in-the-loop review, then set the escalation threshold by gut feel. Headroom measures the part nobody instruments — the reviewer: where judgment degrades under load, and how much reviewers disagree about what’s even risky — turning “is our oversight any good?” into a measured policy, not a guess.
Selected work
AI Objectives Institute — Lead AI Engineer. Built — solo — the multimodal WhatsApp intake agent (voice + text) for Talk to the City, an open-source tool used in national-level democratic consultations; deployed across 55 universities at 10–20K messages/day.
OperatorLock — Risk-control infrastructure that makes a trader's own rules unbreakable: hard, non-overridable gates enforced server-side, where other platforms only warn.
Generative Alpha — AI Developer & Data Engineer. Built a decade-scale SEC-filing ingestion pipeline and fine-tuned GPT-4 / LLAMA2 into vertical financial-research agents.
Beyond engineering
In high school, I founded and led Kids of Hope — a sports camp for 400 Syrian refugee children in Kilis, Turkey, after UNICEF told me I was too young to run one. Photos →