Luka Lukić Email & Phone Number

Germany

San Jose, California, US

Eindhoven, North Brabant, NL

• lead:
• Responsible for ML/DL/CV core algorithm and architecture design, advances, and best practices contributing to multiple verticals (e.g., dashcams, LiDAR, aerial, satellite imagery, probe) for map attribute detection and.
• Ramped up a team and designed the algorithm architecture for the map feature extraction capability from dashcams as a source
• Established and owned the cross-team umbrella of initiatives for the label-efficient machine learning development that led to multi-million Euros per year in savings and cut model development times

Eschborn, Hessen, DE

• lead:
• Developing the full LiDAR perception stack for ADAS and AD for European OEMs (L3):- developed deep learning object detection and semantic segmentation, geometric object detection,multi-object tracking, ego-motion.
• Imaging radar PoC: deep learning object detection, geometric object detection and multi-object tracking
• SLAM PoC: LiDAR-only mapping and localizationacademia and outsourcing:
• established partnerships and projects with 3 outsourcing companies
• co-supervised 3 interns and 2 M.Sc. studentsleading 3 engineering groups: Algorithms, Imaging Radar and Systems and Toolsadvising the executive management on strategic investments and the company approach to.

Eschborn, Hessen, DE

• lead:
• Developing the full LiDAR perception stack for ADAS and AD for European OEMs (L3):- developed deep learning object detection and semantic segmentation, geometric object detection,multi-object tracking, ego-motion.
• established partnerships and projects with 2 outsourcing companies
• co-supervised 3 interns and 2 M.Sc. studentsleading 2 engineering groups: Algorithms and Systems and Toolsadvising the executive management on strategic investments and the company approach to ADASperception

Eschborn, Hessen, DE

• lead:production projects:
• Surround-view monitoring (SVM) production system (4 fisheye cameras): delivered the full online camera calibration algorithm necessary for the top-view stitching
• Park collision assist rear (PCA-R) production system (backward-looking fisheye camera): delivered the multi-object tracker to the exceeding the tracking performance of the previous baseline by 8%
• Rear-side view production system (RSV) (2 mirror replacement cameras): · developed and adapted the multi-object tracker, optical flow and object detection and set up an in-vehicle demo with the object detection which.
• Forward-looking camera (scalable platform from L2 to L4): development of optical flow, ego-motion, online camera calibration, detection of independently movingobjects, head-tail-lamp detection (HBA), multi-object.
• developed a full-scale data annotation tool focused on head-tail-lamp detection (HBA) with a deep-learning superresolution algorithm integrated for more accessible annotation of small objects

Eschborn, Hessen, DE

• lead:Forward-looking camera (scalable platform from L2 to L4):
• improved the stability and speed of the ego-motion algorithm in C
• improved the stability and speed of the optical flow algorithm in C++/CUDA
• improved the stability and speed of the multi-object tracker algorithm in C
• prototyping and improvements of the road freespace detection and tracking algorithm and porting it to C
• prototyping and improvements of the online calibration algorithm and porting it to C

Eschborn, Hessen, DE

• lead and developed:Forward-looking camera (scalable platform from L2 to L4):
• developed the camera calibration tool for intrinsic and extrinsic calibration in C++
• developed tools for ego-motion SIL tests in C++
• ported the tracker algorithm to Ctooling development:
• developed of the auto-tuning module for the tracker parameters in MATLAB
• ported the camera calibration tool to C++

Eschborn, Hessen, DE

• developed: Forward-looking camera (scalable platform from L2 to L4):
• developed the multi-object tracking algorithm in C++ and MATLAB
• improved the capabilities of the existing time-to-collision code in C++ and MATLAB
• improved the pedestrian proposal detector to be ~500x more efficient with a minimal loss of performance in C++
• prototyped the road freespace detection algorithm in C++ tooling development:
• developed the tools for the time-to-collision and AEB SIL tests in MATLAB

Lisboa, Lisboa, PT

• key areas: cognitive robotics, computer vision, machine learningachievements:
• refactored the existing code for visually-guided grasping for the humanoid robot iCub in C++
• wrote and submitted 3 EU-funded research project proposals
• supervised 2 master students

Lisboa, Lisboa, PT

• key areas:cognitive robotics, computer vision, machine learningachievements:
• developed two novel algorithms for motor-primed visual saliency in C++
• developed several novel improvements to the visuomotor controller for the humanoid robot iCub in C++

Lausanne, Vaud, CH

• key areas: cognitive robotics, computer vision, machine learning, statistical data analysisachievements:
• developed a novel visuomotor controller for the humanoid iCub in C++, learned on the human data in MATLAB
• performed analysis on the mocap data in MATLAB, which lead to novel insights on the human eye-arm-hand coordination

CS

• key areas:cognitive robotics, machine learning, fuzzy-control, modelingachievements:
• developed a simulator for a 52-DoF humanoid robot with a virtual-reality interface in MATLAB/Simulink
• implemented motor controllers based on fuzzy logic and artificial neural networks in MATLAB and C++

Doctor Of Philosophy (Phd), Electrical Engineering

Doctor Of Philosophy (Phd), Electrical And Computer Engineering

Master'S Degree, Mechatronics, Robotics, And Automation Engineering