Master’s Thesis Student – Computer Vision for E-commerce

CISTUS AI is a Stockholm-based AI consultancy founded by Onur Yüksel, specializing in computer vision and end-to-end AI solutions. The company also develops web crawling pipelines with agentic capabilities for automated data harvesting.Taking expertise from a 2-year medical computer vision project, CISTUS is exploring e-commerce, with a goal of training a self supervised model that is knowleadgeable of a large fashion retail dataset. The self supervised models in this domain can be used for a variety of down-stream tasks such as image search, classification (highly skilled) or visual item recommendation.The purpose of the thesis project is to optimize, evaluate and discuss the effect of selected components of training pipeline on model's learning.How does the hands-on model development look in this project?Often in computer vision projects we design a "recipe" that describes the process of creating a model that is capable of intended tasks. The recipe describes;how the data is handled during training; what type of augmentations are good for this domain,what type of model architecture is used,designing training mechanics, schedulers, fixing and searching hyper parameters,and evaluation approach.Finding the right recipe is a solid asset for building powerful models.Where does a master thesis fit in?If you think of the recipe above, we should aim for a controlled experimentation with one or multiple components of the recipe. This gives us the flexibility with the scope of the thesis as well. Academic relevance?I believe if we (1) discuss the data/distribution the right way, (2) discuss the approach for designing such experiment, it will be highly relevant.Further goalsHopefully a paper (apart from the thesis), perhaps a workshop or a presentation if we have good results.Who are you?Preferably a master's student in Computer Science or related fields. Entrepreneurial mindset is always appreciated. It would be great if you are ambitious for paper publishing as well. Python, PyTorch, ML/CV knowledge is required.Project Management/LocationWeekly check-in, progress update.Weekly or bi-weekly meetups at a coworking place would be great. Start date : Flexible, but preferably in Jan 2026.Number of students : 1