Engin Oksuz

I work on perception and localization for machines that operate off the map: unstructured terrain and GNSS-degraded air. C++, ROS 2, and Jetson. I do not build the vehicle.

I am a robotics software engineer at HAVELSAN — Robotics and Autonomous Systems. Employer UGV and UAV programs are confidential. I own perception and localization pieces — I do not lead the platform.

Public work is listed under Projects. SIU 2026 is the field paper. Open-benchmark air and ground localization pages are for reproducible eval without employer data. An uncertainty-aware off-road semantics dataset project is in progress; it is not the employer stack.

What I can show

• Field UAV map-aided localization vs RTK — SIU 2026
• Public UAV geo-localization — OrthoLoC + AerialVL
• Public ground localization — MulRan / UrbanNav, GPS on vs denied
• Perception demos (follow-me, etc.) under Projects; not a substitute for the localization pages above

Focus

• Visual and LiDAR–inertial localization, including GNSS-denied stretches
• Multi-modal sensor fusion (camera, LiDAR, IMU, GNSS when available)
• Onboard vision: segmentation and tracking, TensorRT on Jetson
• Map-aided UAV localization against satellite / orthophoto geodata (SIU 2026, public benchmarks)

Not here

Employer platform names, traversability / risk / world-model internals, and papers still under review.

Skills

Languages: C++, Python, Bash
Robotics: ROS 1/2, Docker, Ubuntu, Git
Perception / ML: PyTorch, OpenCV, PCL, Open3D, DINOv2, SAM2, YOLO, ONNX Runtime, TensorRT, CUDA
Estimation: LiDAR–inertial SLAM, EKF / pose-graph fusion, visual geo-localization
Hardware: Jetson AGX/NX, Velodyne / Robosense, industrial cameras, Applanix / SBG, SmartMicro radar