Mikel Broström
Technical Lead at Zeekr Technology Europe
Gothenburg, Sweden
Summary
🤩
RockstarMikel Broström is a Technical Lead based in Gothenburg with a decade of experience building production-grade vision AI and LLM solutions for automotive and edge systems. He has driven end-to-end projects from data collection to deployment—shipping in-cabin monitoring, multi-object tracking, and fully on-device STT-LLM-TTS pipelines optimized for Qualcomm NPUs. At Zeekr he now focuses on multi-agent orchestration, memory systems, and prompt optimization while balancing hands-on ML engineering and DevOps. His open-source work (e.g., BoxMOT) highlights practical improvements to tracking evaluation and CI automation, reflecting a pragmatic approach to reproducible research. Earlier roles span applied research in 3D tracking for autonomous driving and even hardware-focused quantum lab work, demonstrating uncommon breadth across software, ML, and hardware. He combines creative problem solving with a bias for production impact, turning advanced models into dependable features in constrained environments.
10 years of coding experience
5 years of employment as a software developer
Stackoverflow
2,883reputation
283kreached
123answers
20questions
python
top-5%
Github Skills (14)
multi-object-tracking10
python10
cicd10
deeplearning-ai9
machine-learning9
deep-learning9
tensorrt7
object-detection6
yolov86
computer-vision6
pytorch6
yolov56
opencv6
pytorch-lightning6
Programming languages (8)
Github contributions (5)
mikel-brostrom/boxmot
Jun 2021 - Jan 2023
BoxMOT: pluggable SOTA tracking modules for segmentation, object detection and pose estimation models
Role in this project:
ML Engineer & DevOps Engineer
ML Engineer & DevOps EngineerContributions:12 releases, 7 reviews, 191 commits in 1 year 7 months
Contributions summary:Mikel automated the MOT16 evaluation process by creating an evaluation script and integrating it into the CI workflow. They focused on modifying the tracking model, improving the codebase, and fixing bugs. The user also added functionalities such as displaying class and confidence, and contributed to file path updates.
tracking-by-detectionosnetdeep-learningobject-detectioncomputer-vision
mikel-brostrom/test
Oct 2020 - Jan 2024
Contributions:1 PR, 2 pushes, 1 branch in 3 years 3 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.