Where I've worked
- Predicted chaotic dynamical systems over long time horizons with less than 5% error using neural networks.
- Developed and evaluated 5+ neural architectures, including LSTM, GRU, and Reservoir Computing.
- Evaluated models using divergence time, F-statistics, R², and other statistical measures.
- Applied neural forecasting methods to chaotic blood-dynamics models and drafted a research paper presenting the findings.
- Developed a custom self-organizing map (SOM) to project high-dimensional data into a two-dimensional topological representation.
- Evaluated SOM performance using high-dimensional topological structures generated with parametric equations.
- Built the model in Google Colab using scikit-learn, SciPy, and NumPy.
- Created visualizations of data and experimental results using Matplotlib.
Where I've studied
Codeforces Difficulty Prediction Model
Built a full-stack web application that predicts Codeforces problem difficulty with less than 10% error compared with actual ratings. Applied statistical regression to historical contest data, developed a Flask backend with a Supabase SQL database, and automated data collection through the Codeforces API. The application attracted approximately 100 unique users.
Logicdle
Built, tested, and deployed a full-stack application in under 40 hours using Claude Code. Implemented a user submission pipeline, image storage and retrieval, and leaderboard functionality. Reviewed the code to ensure sensitive information, including passwords and database keys, was handled securely.
AI Coliseum — 5th Place
Developed an autonomous game-playing agent for the AI Coliseum 2024 competition. Designed a symmetry-based exploration strategy that used bot dispersion to map unknown environments efficiently, implemented inter-agent communication to share discovered regions, and developed a navigation algorithm to avoid walls and environmental hazards.
I'm interested in machine learning, theoretical computer science, algorithms, and mathematics. If you are working on an interesting research problem or building something exciting, feel free to reach out.
Send a message