M

ML

🛒 AI chat & product/category summaries in Amazon shopping, powered by the latest LLMs

The main project for OctoMY™ - Ready-to-run robot software

Practical tasks on Deep Learning (DL) and Neural Networks (NN). 🤖

VERA (Video Extraction and Recognition Analysis) is a fully on-premises video OCR pipeline for extracting and recognizing text (license plates, street signs, billboards) from dashcam/driving footage. Built with PaddleOCR 3.x, OpenCV, FastAPI, PostgreSQL, and MinIO, with a PySide6 desktop GUI. Consists of two components: vera-engine (OCR pipeline, backend API, CLI) and vera-desktop (GUI layer). No cloud inference, all processing runs locally.

FairLiebt Matching-Service — Python 3.12 / FastAPI (regelbasiert Phase 1 + ML Phase 2)

A dual-channel ML-based spelling correction engine for Movies and TV search, trained on TMDB vocabulary.

Fundamental theory and practice in Data Science (DS). 🧮

CVerify is a production-ready AI recruitment platform that analyses CVs through a 3-phase ML pipeline, recommends matching jobs from 20 000+ real job postings, and provides a full recruiter/candidate/admin web interface with an integrated AI chatbot.

My personal webpage

Simple ML Project using regression models and KMeans clustering to predict Y from A and B, classify results, and expose predictions through a FastAPI API.

FastAPI-based microservice that predicts the risk level of a user session https://user-risk-detection-api.vercel.app

Predicts the risk level of a user session using behavioral signals

In real time inference simulator using the MVTec dataset, aproach via DINOv2 and Claude Sonnet 3.5 (Amazon Bedrock's API)

Slimme huishoudboekje-app met AI-gestuurde categorisering. Importeer je ING/Revolut transacties, krijg automatisch inzicht in je uitgaven en beheer je budget met NIBUD-referenties. Privacy-first: alle ML draait lokaal.

Volledige ge amp/vibe coded met Claude Code!

The AskAnna CLI offers a command-line interface to the AskAnna platform. With the Python SDK you can run AskAnna functions directly from your Python script. The CLI & Python SDK simplifies the communication with the AskAnna platform.

FOMC rate decision predictor using ML bronze->silver->gold methods

A production-oriented Machine Learning pipeline that predicts whether an active user session will result in a purchase.

Model: XGBoost Classifier optimized for class imbalance.

Performance: ROC AUC 0.936 | F1-score 0.71 (at 0.30 threshold).

Key Features: Reproducible environment (uv), modular CLI for training/inference, leakage-free preprocessing, and SHAP interpretability analysis.

Data: UCI Online Shoppers Purchasing Intention Dataset.

Tech Stack: Python, XGBoost, Scikit-learn, Pandas, SHAP.

A lightweight, AI-powered Applicant Tracking System (ATS) that parses PDF resumes using local LLM inference. Built for personal/small team use with a modern, minimal interface.