retraining

Graph Transforms to Quantize and Retrain Deep Neural Nets in TensorFlow

A image classifier that classifies whether a galaxy is spiral, elliptical, or irregular. Built using Tensorflow v2 and Google's inception v3 model.

In this repo, we would see how to handle streaming data using "creme" library

An anomaly detection model for HDFS_v1 log dataset. Kafka to simulate real time data streaming and model retraining on new unseen data.

End-to-end adaptive ML system with real-time prediction, monitoring, drift detection, and automated retraining using FastAPI and Streamlit

A self-supervised learning framework based on DINOv3 (Vision Transformer) for custom dataset retraining. This project implements training, evaluation, and visualization tools with flexible configuration options.

AI Engineer & Full‑Stack Developer building production LLMs, computer vision and forecasting systems with Python, FastAPI and Next.js for real business impact.

Enhancing electricity price forecasting accuracy using a hybrid model combining GRU and XGBoost with detection-informed retraining for concept drift.

Retrain an existing COCO model, and output it with Tensorflow Lite. Shows how to do quantization so you can use it with for example Coral USB Accelerator.

Retraining of LLM Models using LoRA and QLoRA

End-to-End ML Deployment Platform with Drift Detection

A Machine UnLearning pipeline for Detectron2 (RetinaNet), designed to surgically remove neural backdoors from space debris streak detectors without degrading legitimate detection. Built for the ESA 'Secure Your AI' Kaggle Competition.

Python project to show why retraining can be harder than training in a changing environment

Vision Transformer Level 2 MLOps Deployment Scalable ONNX-based ViT pipeline with FastAPI, MLflow, and event-based retraining.

End-to-end, production-aware fraud detection system with real-time inference, delayed labels, explainability, drift monitoring, retraining, and full audit governance.

Predicts 30-day hospital readmission risk using explainable ML, with real-time monitoring, drift detection, and automated retraining pipelines.

Scheduled ML retraining pipeline — DVC data versioning, MLflow experiment tracking, champion/challenger gating, and full audit trail. Auto-retrains weekly via GitHub Actions. Project 2/10 of MLOps portfolio.

This template uses voting for combining classifiers and it shows how to use the backtester with retraining option.

Dokumentation der Projektarbeit im Rahmen der Umschulung Fachinformatiker:in Anwendungsentwicklung