About

I make AI systems
behave in production.

E

I'm Emre — a software engineer based in Turkey. My background is in machine learning and data systems; the last few years have been specifically about making LLMs behave predictably inside real products.

I spent 5+ years at GameTheory Games as a data scientist, which meant shipping ML models that actually had to work — player behavior forecasting, monetization analytics, real-time pipelines. That's where I stopped thinking about models and started thinking about systems.

"I don't list every technology I've touched.
I list what I can defend in a technical conversation."

What I Believe
01
LLM is a component, not a product.
The value lives in the system around it — the state machine, the validation layer, the human review queue. I design those.
02
Every decision must be traceable.
If you can't explain why your AI did something, you don't have a system — you have a black box with a chat interface.
03
Production ≠ demo.
I've watched beautiful demos collapse on day one. I build for the Monday after launch, not the Friday before it.
04
Constraints are the product.
The guardrails, the fallbacks, the policy layers — those aren't limitations. They're the reason the system earns trust.
Experience
2019–
Data Scientist
GameTheory Games
5+ years building ML systems for a live game: player behavior modeling, churn prediction, in-app purchase forecasting, real-time analytics pipelines. Learned what production actually means.
2023
Systems Shift
Independent
First LLM integrations. Moved from model-centric to systems-centric thinking — FSM-based orchestration, policy validation, agentic workflows. The defining change in how I work.
2024–25
AI Systems Engineer
Independent
Ticket-Mind: FSM + LLM triage agent. HermesReplier: AI-powered Gmail automation. Churn retention system that goes beyond prediction into decision logic.