I work as part of the Data Science/Engineering team providing business insights to companies by analyzing their customer data. Main technologies that i use include Python, Google Cloud Platform, Adobe Analytics, and SQL.

Spent multiple semesters as teaching assistant in courses ranging from introductory programming to symbolic AI.

Developed a prototype app using the Ionic framework. The app served nutrition information and relevant offers from nearby stores using the FoodInfo API.

GitHub Repo
I presented a unified framework for implementing adversarial search algorithms, e.g. MiniMax and MCTS, with the aim of facilitating research, education, and open source collaboration in the field. A side effect of the framework is that it introduces an 'agent space' that allows for automated search for strong adversarial agents.

GitHub Repo
We trained the You Only Look Once object detection model to detect the helmet usage of motorcycle and scooter users in a dataset recorded in developing countries. A positional encoding was used to deal with the many classes resulting from up to five people on each vehicle, each either wearing or not wearing a helmet.

GitHub Repo
We developed a product recommendation system for Pandora jewellery using two approaches; an autoencoder and a triplet ranking network. This was paired with an object extraction model from Facebook Research, which resulted in a fully functional pipeline from input image to recommendations.

Link to Demo
We built a webscraper to scrape the contents of all courses listed on The DTU Course Base. Using a TF-IDF analysis we found the words that best described each course, and used these shared words to form a large network of courses. The graph can be explored in a 3D space, and shows a high degree of clustering for subjectively similar courses.

By making a Q-learning agent train against itself in a spaceship dogfighting game we programmed from scratch, we managed to raise the agents' performance to a level where it regularly beat human opponents.