Håkon Hukkelås
Software Architect at Maritime Robotics AS
Trondheim, Trøndelag, Norway
Summary
🤩
Rockstar🎓
Top SchoolHåkon Hukkelås is a Software Architect based in Trondheim with 11 years of experience building production-ready ML and computer vision systems. He holds an MSc and is pursuing a PhD in Computer Science from NTNU, and has combined academic depth with industry impact at Maritime Robotics and Cisco. His work spans low-footprint, CPU-friendly deep learning (face and keyword detection) to generative models for privacy-preserving image anonymization, contributing concrete WGAN/ResNet implementations and FID evaluation improvements on GitHub. Håkon moves between hands-on model engineering and system-level architecture, shipping end-to-end solutions from training loops to backend integration. Colleagues rely on him for reducing false positives in constrained environments and for translating research prototypes into robust products. He’s particularly interested in generative models and autonomy, pairing research curiosity with pragmatic delivery.
11 years of coding experience
2 years of employment as a software developer
University of California, San DiegoNorwegian University of Science and Technology
English, Norwegian
Github Skills (15)
computer-vision10
machine-learning10
pytorch10
cgan10
cyclegan10
python10
machine-learning-models10
dcgan10
fine-tuning9
resnet9
deep-learning9
tensorboard9
cuda8
data-augmentation7
object-detection5
Programming languages (2)
Github contributions (5)
hukkelas/DeepPrivacy
Oct 2018 - Jan 2023
DeepPrivacy: A Generative Adversarial Network for Face Anonymization
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:241 commits, 11 PRs, 32 pushes in 4 years 3 months
Contributions summary:Håkon's commits center around the development of a WGAN-based model for image anonymization using a ResNet architecture. The user implemented the necessary layers, loss calculations (Wasserstein loss, gradient penalty, epsilon penalty), and a training loop within `train.py`. The commits show modifications to the InceptionV3-based FID score implementation for evaluation, indicating work with evaluating the model. Furthermore, the user refactored core components and integrated both the generator and discriminator for the final model.
deepfakerobustnessadversarialgenerative-adversarial-networkanonymization
HoneyDrive/HoneyDrive
Mar 2016 - Apr 2016
Contributions:65 commits, 36 PRs, 47 pushes in 1 month
ultimatepython
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.