Founder of HisseVergiBeyan and senior robotics software engineer at Aeolus Robotics focused on household manipulation for mobile service robots. Lately I've been extending our reactive action processor with new manipulation skills, stewarding fetch-and-delivery, disinfection, door-manipulation, and elevator pipelines, and playing a significant role in developing custom grippers (e.g. a novel UVC disinfection device) tailored to those deployments.

Right now my core emphasis is learning-enabled manipulation with large models 🤖. I built the online planning and execution pipeline that exposes trajectory rollouts and reward signals for on-robot learning, integrated hardware-accelerated planners so expert trajectories arrive fast enough to iterate, and designed the planning skill-benchmark suite—covering skills like button-pushing, grab, dispose, and door-sliding—to keep behavior regressions visible.

Before Aeolus, I wrote guidance, navigation, and control software for DJI Matrice 100 platforms at MilSOFT. I implemented the autonomy stack, spun up a high-fidelity Gazebo simulation and hardware-in-the-loop lab ✈️, and connected embedded controllers with mission software so flight-readiness checks could run overnight across swarms of test vehicles.

• Built HisseVergiBeyan into a detailed tax-reporting assistant for 🇹🇷 investors trading on foreign stock exchanges—it's humming along even without active development these days.
• Automating profit and loss breakdowns with verbose explanations of splits, delistings, and ₺ inflation adjustments so annual filings stay accurate.

• Extending manipulation skill coverage across fetch-and-delivery, disinfection, door-manipulation, and elevator missions with custom grippers.
• Maintaining articulated collision object libraries and telemetry-fed online replanning pipelines for humanoid deployments.
• Stewarding skill-level benchmarking infrastructure—covering button-push, grab, dispose, and door-slide behaviors—to keep regression detection grounded in real deployments.
• 👐 Contributed upstream improvements to notable open-source software, e.g. MoveIt, MoveIt Tutorials, and rrt-algorithms.

• LinkedIn
• Email

Check out my open-source work below—especially hector-moveit, sjtu-drone, and the dji-m100-ros organization that houses my Matrice 100 autonomy modules.

These days I'm applying vision-language-action pipelines and large behavior models to Aeo (a humanoid robot mounted on a mobile base), wiring up logging, visualization, and curation workflows that make large-scale learning iterations feasible. Always happy to compare notes with anyone exploring dependable robotic behavior at scale 🤝