I joined StarApps through a recommendation from my previous student job at Realo. StarApps is a venture studio specializing in creating and funding multiple startups under a single team. The company exclusively focuses on SaaS startups, leveraging its extensive experience in this domain.
My role at StarApps involves handling the technical aspects of a startup named Track, a private equity investment tracker. Building a startup or an MVP encompasses a diverse range of responsibilities, including full-stack development, designing AI solutions, establishing infrastructure and architecture, and deploying systems from the ground up to create production-ready solutions. As such, my role isn’t confined to one specific area but rather spans multiple disciplines.
Currently, the technical team consists of three people, with me being the only one having full focus on Track. This environment offers invaluable learning opportunities, allowing me to grow while collaborating with brilliant and successful individuals.
During this internship, I built upon my previous work at Realo from the year before. Realo is a real estate platform that not only operates as an agent but also aggregates the entire Belgian housing market from various real estate websites. Capturing such vast amounts of data requires a robust IT infrastructure to derive actionable insights.
My task was to develop a neural network to predict house prices at the city level. This project presented several challenges, including forecasting time series, managing limited data with a high-dimensional feature space, and working with thousands of small time series. While I initially fully implemented and optimized an LSTM approach to address these issues, a random forest model (using XGBoost) ultimately delivered superior results.
During this internship, I developed data cleaning pipelines and meticulously prepared the data for future tasks, including steps such as outlier removal, imputation, and more. While many data scientists find this process tedious, I genuinely enjoy the challenge and precision it requires.
During this internship, I developed a Python module to replicate a linear programming solution originally implemented in C++, making it more accessible and user-friendly for a broader audience at ArcelorMittal