TurboMino: Tetris AI Bot with PPO + CNN + RoPETurboMino: Tetris AI Bot with PPO + CNN + RoPE

Apr 2026

In progress

Tetris agent trained with PPO (Stable-Baselines3) on a custom Gymnasium environment. Uses CNN with Rotary Position Embeddings (RoPE), bit-packed uint32 board representation, exhaustive placement search via BFS, and interactive Pygame visualization.

AI

Jupyter

NumPy

Pandas

Pygame

Python

PyTorch

Reinforcement Learning

AI

Jupyter

NumPy

Pandas

Pygame

Python

PyTorch

Reinforcement Learning

TFM: Reward Machines on RLTFM: Reward Machines on RL

Apr 2026

In progress

Master's thesis implementing tabular Q-Learning augmented with Reward Machines (reward automata) and Counterfactual Reward Machines (CRM). Environments: Taxi-v3, MultiTaxi (2 passengers), MiniGrid-DoorKey. Benchmarks 5 variants (no RM, RM only, RM+CRM, improved versions).

AI

Jupyter

Matplot

NumPy

Pygame

Python

Reinforcement Learning

AI

Jupyter

Matplot

NumPy

Pygame

Python

Reinforcement Learning

String artString art

Feb 2025

Finished

From pixels to lines. This algorithm transforms raw numbers into patterns that resemble art. Start from one point, evaluate all possible lines, and draw the one that improves the image the most. Repeat until you're done!

Git

JavaScript

Next.js

React

Tailwind CSS

TypeScript

Git

JavaScript

Next.js

React

Tailwind CSS

TypeScript

TFG: Visual Document UnderstandingTFG: Visual Document Understanding

Mar 2025

Finished

Bachelor's thesis evaluating Visual Document Understanding methods for extracting structured information from documents. Compares Azure OCR + LLM, custom fine-tuned OCR, and DONUT (Document Understanding Transformer) with PyTorch Lightning. Deployed with Docker.

AI

Docker

HuggingFace

Jupyter

Matplot

NumPy

Pandas

Plotly

Python

PyTorch

AI

Docker

HuggingFace

Jupyter

Matplot

NumPy

Pandas

Plotly

Python

PyTorch

Tetris Solver: GA & Simulated AnnealingTetris Solver: GA & Simulated Annealing

Sep 2025

Finished

Modern Tetris implementation in Unity solved with two metaheuristics: Genetic Algorithm (GA) and Simulated Annealing (SA). Each solution is a sequence of piece placements evaluated by a fitness function that minimizes board space. Includes Python analysis tools and a research paper.

AI

C Sharp

Jupyter

LaTeX

Matplot

Pandas

Plotly

Python

Unity

AI

C Sharp

Jupyter

LaTeX

Matplot

Pandas

Plotly

Python

Unity