Experience
Developing latent-space video world models, scaling DINO to the billions-parameters class, learning self-supervised learning of object-centric representations, and vision-language alignment.
Research internship during my PhD, worked on semantic image compression with vector-quantized autoencoders and DINO saliency maps. Presented the paper “Variable Rate Allocation for Vector-Quantized Autoencoders” at ICASSP 2023.
Research internship during my PhD, worked on explainable detection of DeepFake videos. Presented the paper “Quantitative Metrics for Evaluating Explanations of Video DeepFake Detectors” at BMVC 2022.
Research internship for my Master’s thesis, worked on generative models for virtual try-on: combining one’s own photograph with the stock picture of a fashion garment. I gave a talk about this project at the NLDL workshop 2019 in Tromsø, Norway. The full thesis available at this page.
Forecasting sales to cut sourcing costs and reduce food waste at the first Malaysian startup accepted into Y Combinator. Helped to implement route optimization algorithms for the delivery riders.
Education
Thesis on structured representations for model explainability in deep learning and computer vision. Supervised by Prof. Hossein Azizpour at the division of Robotics, Perception and Learning. Teacher assistant for the courses of “Artificial Intelligence”, “Deep Learning”, and “Deep Learning Advanced”. System administrator of in-house compute cluster of the department (100+ GPUs, 70+ users). Supervised the master’s theses of 3 students.
Thesis on generative models for virtual try-on, supervised by Prof. Josephine Sullivan and Urs Bergmann at Zalando Research. Teacher assistant for the courses of “Artificial Intelligence” and “Image Analysis and Computer Vision”. Represented KTH in the university rowing team and competed in the Swedish Championships.
Graduated with honors (110L/110). Internship at CASY robotics lab on optimizing running parameters of motors for quadcopter drones. Thesis on smart home automation with physical sensors on Raspberry Pi devices distributed over the network,