I built a deep learning model to classify brain tumors on MRI scans. By achieving high accuracy (up to 99%) and visualizing results with Grad-CAM, this project demonstrates how AI can contribute to early detection and diagnosis in healthcare.

I built Bondora, an app that helps people improve their relationships through understanding the underlying behavioral patterns.

I built a 2D Platformer adventure game using Pygame, focusing on realistic physics and game mechanics

We built VisionSF, an immersive event planner in collaboration with the city of San Francisco on the Apple Vision Pro plattform.

During my time in San Francisco, we created World Schedule. A Meeting Planer that connects people over different timezones. It immediately played out to be useful.

I recreated Andrej Karpathy's NanoGPT from scratch to dive deep into the mechanics of Large Language Models (LLMs) and Transformers. This hands-on project allowed me to understand how LLMs generate human-like speech on the most fundamental level.

I built an autonomous Agent to land the Lunar Lander using Gymnasium API and Stable Baselines3, implementing the DQN algorithm.

I did a project on the Prisoner's Dilemma to analyze which strategies work best in game theory. This project can be applicable in real life strategic situations.

I applied GradCAM on the classical MNIST Digit Recognition Problem to visualize most influential features for model decision.

I implemented the Backpropagation algorithm from scratch to understand how neural networks work on the most fundamental level.

I developed a strategy game with a team of four people using Maven.

I have some notebooks on the Hands-on ML book.

{Tech: Europe} Berlin AI Hackathon. Project: Development of an AI Agent that can tutor children in math.

USS Hornet Defense Tech Hackathon. Project: Development of a drone pilot detection system

I am building an interface between the highly realistic MARL Simulator ARGoS and the MARL framework PettingZoo. This will allow for more realistic simulations of multi-agent systems in swarm robotics.

As AI models became too complex for humans to comprehend, it is essential that users remain having trust in these models. Therefore, Explainable AI (XAI) is an important research field. My study compares two general approaches -Black-Box & White-Box methods- for explaining how Deep Convolutional Neural Networks form decisions. The specific methods used were LIME & GradCAM.