I'm a machine learning engineer, software developer, and robotics researcher based in Türkiye. For the past fifteen years, I've helped research labs and startups turn ambitious ideas into working systems from learning algorithms to physical prototypes.
From the first dataset to the final fabricated part, I work across the full stack of modern engineering — bridging machine learning, software, and robotics into systems that actually work in the real world.
End-to-end machine learning — from raw, messy data to deployed models that hold up under real production loads. I focus on systems that are interpretable, efficient, and genuinely useful.
Core stackHand-crafted, scalable codebases that age well. Whether it's a research prototype or a production service, I write software that's readable on Monday and maintainable a year later.
Core stackClosing the loop between perception, planning, and control. I build robotic systems that move with intent — from low-level dynamics to high-level autonomy.
Core stackFrom CAD model to physical part. I take ideas through 3D design, simulation, and additive manufacturing — so the things I prototype actually exist in the room with you.
Core stackTurning open questions into demonstrable results. I read the papers, run the experiments, write the report, and build the prototype that proves whether the idea is actually worth pursuing.
Core stackModels are only useful if they survive contact with production. I set up pipelines, monitoring, and the boring infrastructure that keeps trained models running, observable, and continuously improving.
Core stackA small selection of recent projects across machine learning, robotics, and software engineering — each one taught me something new about working systems.
A deep learning pipeline that classifies surface defects on automotive parts in under 40ms per frame. Deployed on edge GPUs across three production lines, cutting manual inspection time by 72%.
Read the case studyA reinforcement learning policy for a collaborative robot arm, trained in simulation and transferred to hardware. Achieves 94% grasp success across 30+ unseen object geometries.
View projectA carbon-composite quadcopter frame designed from scratch in CAD, validated through finite element analysis, and produced via SLS printing. 31% lighter than the baseline at equal stiffness.
Visit projectA modular C++ motion planning library for mobile robots, with Python bindings and full ROS 2 integration. Now used by three research groups and listed in 12 published papers.
See the repoFour ways to collaborate, depending on where you are in your engineering journey. Every engagement starts with a free 30-minute call — no pressure, no sales pitch.
For teams looking to apply machine learning to a real-world problem — from feasibility study to a trained, evaluated, production-ready model.
Scalable backends, internal tools, scientific software, and APIs. Hand-written in Python or C++, depending on whether you need clarity or performance — or both.
Perception, planning, and control for mobile robots and manipulators. Long-term partnerships where I embed with your team and ship alongside engineering.
A focused engagement to de-risk an idea before you commit. Useful for grant applications, technical due diligence, or right before a major build.
Research projects, client engagements, and the occasional weekend experiment that turned into a real published paper.
Energy Sector · Ankara
Autonomous Vehicles · Munich
Research · ETH Collaboration
FinTech · Amsterdam
MedTech · Istanbul
IEEE ICRA · Co-authored
Emre is the rare engineer who can move from a research paper to a deployed system without losing rigour on either end. He doesn't just hand off models — he integrates with the team, asks the sharp questions, and somehow keeps the whole process calm. Our robotics stack is genuinely better because of him.
I take on a small handful of engagements each quarter so I can give every one of them real attention. If you're thinking about an ML system, a robotics prototype, or a piece of software that has to actually work — drop me a note. I usually reply within a day.
