Tymon Masiarek

About Me

Now

Right now, I’m doing small freelance projects on the side, reading advanced AI/ML books to deepen my theoretical base, and actively developing my main project - a Security Operations AI Assistant powered by RAG and LLM fine-tuning.

Technologies I Use

Projects

Project 1 - Security Operations AI Assistant (RAG + Whisper + Fine-tuned LLaMA 4)

Developed an AI assistant for security operations using Retrieval-Augmented Generation (RAG). Integrated Whisper for speech-to-text, embeddings with ChromaDB, and fine-tuned LLMs for context-aware decision support. Deployed via FastAPI

Live Demo View Code

Project 2 - Research Paper Implementation - Attention is all you need

Reproduced the Transformer architecture from the “Attention Is All You Need” paper in PyTorch. Implemented encoder–decoder attention from scratch, trained on benchmark data, and validated results against reported performance.

Live Demo View Code

Project 3 - Linear Regression from scratch

Implemented Linear Regression in NumPy and PyTorch, coding gradient descent and backpropagation manually. Ran experiments with learning rates, Ridge regularization, and GD variants, visualizing convergence and performance on Kaggle dataset.

Live Demo View Code

Project 4 - Customer Churn Prediction

Built a machine learning pipeline to predict telecom customer churn. Preprocessed features, trained Logistic Regression, Random Forest, and XGBoost models, and achieved 85% ROC-AUC. Identified top churn drivers and provided actionable business insights.

Live Demo View Code

Project 5 - Handwritten Digit Recognition from Scratch

Implemented a neural network from scratch in NumPy with manual forward and backward propagation, trained on MNIST. Added an interactive demo with a drawing canvas and real-time confidence bar visualization for digits 0–9.

Live Demo View Code