We are seeking a Stockholm-based Master’s student in Mathematics, Finance, CS, or Data Science to undertake a degree project aimed at assessing the risk of rental payment defaults. This project draws inspiration from insurance claim modeling and financial risk assessment, with the goal of predicting the likelihood that a tenant will default on rental payments. The project will involve building models to assess risk, segment users into risk groups, and create automated alerts for potential defaults. The successful candidate will work at our office during the Spring semester, collaborating closely with our data science, product, and development teams.

What You’ll work on

• Research and Model Development: Explore statistical and machine learning approaches to predict the risk of a defaulting payments. You’ll analyze historical rental data, user profiles, and market conditions to develop models.
• Feature Engineering: Identify key features that contribute to default risk.
• Modeling Techniques: Leverage various techniques such as logistic regression, decision trees, gradient boosting, and survival analysis to assess default risk, similar to insurance claim modeling.
• Risk Segmentation: Categorize different risk groups and build a risk scoring system to help Qasa and users make informed decisions about lease agreements.
• Collaborate with the product, financial, and development teams to integrate fraud detection models into our platform.
• Gain hands-on experience in financial risk management, and real-time data processing.
• Communicate your research findings, progress, and final results to both technical and non-technical stakeholders.

You could be a great fit if you (not required)

• Are experienced or interested in learning Python, SQL, Excel in a financial context.
• Are interested in financial risk assessment or insurance modelling.
• Have hands-on experience with statistical or machine learning libraries such as scikit-learn, TensorFlow, PyTorch, etc.
• Excel at applying theoretical and empirical research to solve practical challenges.
• Are passionate about data-driven decision-making and solving real-world problems.
• Can communicate complex research clearly, precisely, and in an actionable way.

What we offer

• An opportunity to work on a high-impact project addressing real-world challenges in the long-term rental market.
• Mentorship and guidance from experienced data scientists, developers, and product/financial experts.
• A collaborative, innovative, and fun working environment.
• Potential future employment opportunities at Qasa.
• Writing a master thesis is fun but for many the last thing they do before graduating. We want you to have a suitable scope and technical support to be able to write and finish your degree.

🎓 Previous Master’s Theses

At Qasa, we’ve had a strong history of collaborating with Master’s students on impactful projects. Here are some of the recent theses completed at Qasa:

• Automated Home Review: Eric Dahlgren developed a system to review home ads before they are published. Using XGBoost and SHAP, an interpretable anomaly detection model was built to automate ad reviews. The model not only scores ads but also provides human-readable explanations to assist decision-making. You can read the thesis here: Enhancement of an Ad Reviewal Process through Interpretable Anomaly Detecting Machine Learning Models
• Rent Prediction: Vasigaran Senthilkumar focused on improving house rental price predictions by incorporating external features such as walk scores and interest rates. The work also introduced methods for handling uncertainty in predictions, including generating prediction intervals and managing uncertain rent data to enhance model adaptability and generalization. You can read the thesis here: Enhancing House Rental Price Prediction Models for the Swedish Market
• Recommendation Systems: Julia Byström created and A/B tested (in production!) two new algorithms to improve tenant rankings for landlords in our “Find tenant” feature. The algorithms were rigorously tested and deployed in production, contributing significantly to our recommendation systems. You can read the thesis here: Probabilistic Weighting and Deferred Acceptance in Reciprocal Recommendations