Nine years of turning hard problems into working systems. I operate at the intersection of machine learning, software engineering, and robotics — and I do it from Istanbul.
I started my engineering career at Istanbul Technical University in 2013, where I studied mechatronics because I couldn't decide between mechanical, electrical, and software engineering — so I chose all three. That same inability to stay in one lane has defined everything since.
After graduating, I joined Atlas Technologies as a software engineer, where I spent two years building internal automation tools and discovered I was most excited when code touched the physical world. That led me to a research position at ETH Zürich's Robotics Systems Lab, where I worked on sim-to-real transfer for legged robots and wrote my first three conference papers.
I've since worked with robotics startups in Munich and Berlin, led ML engineering teams, and returned to Istanbul to build the research side of my practice. Today I work independently, taking on a small number of serious engagements each year — the kind of problems that don't have obvious solutions and require someone willing to read the literature, run the experiments, and build the prototype.
I also write — about engineering, about research, and about the sometimes-uncomfortable gap between the two. And I try to release as much of my work as open source as clients allow.
Industry
Lead researcher on a TÜBİTAK-funded project developing sample-efficient reinforcement learning algorithms for industrial manipulation. Managing two PhD students and collaborating with ETH Zürich on benchmarking.
Built and scaled the ML team from 2 to 8 engineers. Led development of Helios's flagship pick-and-place perception system, now deployed in 6 facilities across Europe. Responsible for the full model lifecycle — research, training, evaluation, and edge deployment.
Developed the perception stack for Pulse's L4 autonomous vehicle platform. Focused on sensor fusion, object detection robustness in adverse weather, and real-time inference at sub-30ms latency on embedded NVIDIA hardware.
Built industrial automation software, internal tooling, and a real-time monitoring dashboard used across 12 manufacturing facilities. Learned that readable code outlasts clever code every time.
Research
Worked on sim-to-real reinforcement learning for legged locomotion. Co-authored three papers on adaptive control policies under Prof. Marco Hutter's supervision. The ANYmal quadruped project remains one of the most rigorous and rewarding technical experiences of my career.
Undergraduate research on nonlinear model predictive control for quad-rotor UAVs. Contributed to the lab's open-source simulation framework and wrote my first technical report — which I'm told was readable, which at the time felt like high praise.
Istanbul Technical University · Istanbul, Turkey
Specialisation in robotics and control systems. Thesis: "Adaptive Reinforcement Learning Policies for Robust Manipulation in Unstructured Environments." Graduated with honours. GPA: 3.82 / 4.0.
Istanbul Technical University · Istanbul, Turkey
Core curriculum spanning mechanical design, electronics, control theory, and software. Final-year capstone: autonomous mobile robot for warehouse logistics. Graduated first in class. GPA: 3.91 / 4.0.
deeplearning.ai · Online
Five-course specialisation covering neural network fundamentals, CNN architectures, sequence models, and structuring ML projects. Completed as supplementary training to complement the robotics-focused M.Sc. curriculum.
ETH Zürich · Zürich, Switzerland
Non-degree research residency at the Robotics Systems Lab. Focused on sample-efficient policy learning for legged systems. Attended graduate seminars in probabilistic robotics and optimal control. Published three conference papers during this period.
A practical snapshot of where I spend my time and what I reach for first on any given project.
The best engineering ideas often come from unexpected places. Here's where I go when I'm not at my desk.
History of science, philosophy of mind, and the occasional novel that makes me late for meetings.
The Bosphorus hills and Belgrad Forest are my favourite debugging environment. No screens required.
Particularly interested in philosophy of mind and what machine cognition can and cannot tell us about intelligence.
Electronic music — ambient and generative mostly. I've been building a modular synthesizer for three years and it's still not done.
I take on a small number of engagements each quarter. If you have a project that fits, I'd love to hear about it.