About
I'm a recent computer science graduate (MS, University of Colorado Boulder) who moved to Copenhagen in Autumn of 2025. I'm currently enrolled in University of Copenhagen and Aalborg University taking classes in Reinforcement Learning and Generative AI. My research interests during my masters were in creating physics informed neural networks, bounding neural network outputs to respect physical constraints. Currently I'm exploring the internal mechanisms of LLMs, particularly how backdoors are encoded in LLMs and how they can be detected, and multimodal learning at Aalborg University under Yan Kyaw Tun.
Research Interests: AI Safety, Formal Verification, Mechanistic Interpretability, Physics-Informed Neural Networks, Reinforcement Learning
Research
Multimodal Learning & Missing Modalities · Aalborg University under Yan Kyaw Tun
Currently researching multimodal learning with missing modalities.
Education
Graduate Coursework · University of Copenhagen
Graduate Coursework · Aalborg University
MS Computer Science · University of Colorado Boulder
BA Computer Science · University of Colorado Boulder
Volunteer
DANSIC — Danish Social Innovation Club
Contributing to initiatives at the intersection of AI ethics and sustainable development.
Coding Pirates
Teaching kids programming fundamentals and hands-on microcontroller projects.
Projects
Predicting Soil Moisture Dynamics with Physics-Informed Neural Networks
Developed a PINN model to predict continuous 1D soil moisture profiles for precision irrigation, integrating the Richards Equation as a physics constraint. The model achieved PDE residuals of 1×10⁻⁷ and was validated against real-world sensor data from 57 stations across 8 global soil moisture networks (RMSE: 0.018 m³/m³). Transfer learning reduced training time by over 50% while enabling 24–72 hour predictions where traditional numerical solvers failed to converge.
Adversarial Backdoor Injection in LLMs
Investigate backdoor persistence in fine-tuned language models. Implemented a modular injection pipeline on Llama 3.2-3B using LoRA with 4-bit quantization, and developed a comprehensive evaluation framework measuring attack success rate, clean accuracy, trigger robustness, and poison efficiency across varying configurations.
BitShogi — A Japanese Chess Game Engine
Built a complete mini shogi ("The Game of Generals") engine in Julia using bitboard representations for efficient move generation and board state management. The project includes a REST API server and a React/TypeScript web frontend playable at bitshogi.com. Featured in the Julia programming language newsletter as a community project.
Urban Management Model for Ride-Sharing
Applied a multi-agent, ant-colony-inspired route-finding simulation to optimize ride-sharing routing in complex urban environments. Built a multithreaded Python simulation modeling pheromone-based pathfinding that adapts in real time to changing traffic conditions across diverse street network topologies.
Curriculum Vitae
A full PDF of my CV is available for download.
Download CV (PDF)