From components to connected systems

I worked on a reinforcement learning project that implemented self–other modeling inspired by ideas from cognitive psychology. The key idea was that an agent infers another agent's goal by putting itself in the other's position, using its own neural network to simulate what it would do and comparing this prediction to the observed behavior. To enable this, I modified a standard RL model so that the loss was backpropagated into the model input rather than the network weights, allowing the agent to iteratively improve its estimate of the other agent's goal during interaction.

My master's thesis: a complete Searchlight Representational Similarity Analysis pipeline for fMRI data, implemented from scratch and integrated into the neuroimaging toolbox QuNeX to map representational structure across the cortical surface.

A cross-platform Flutter app for Android and iOS that supports descriptive experience sampling research, capturing people's moment-to-moment thoughts and feelings through randomized daily prompts.

A mobile app for installing and managing AMR modules that retrofit existing gas and water meters into smart meters, bridging physical metering hardware with cloud infrastructure through a REST API.

Project description coming soon...