OpenMed MLOps
An educational MLOps demonstration that trains a classifier on synthetic patient data with Kubeflow Pipelines and records experiments in MLflow on Kubernetes.
Selected repositories across applied machine learning, language technology, developer tools, and reproducible infrastructure. Explore the source and the ideas behind each project.
16 selected projects across cloud infrastructure, applied AI, and useful software.
An educational MLOps demonstration that trains a classifier on synthetic patient data with Kubeflow Pipelines and records experiments in MLflow on Kubernetes.
Terraform-managed local Docker environments with isolated application and PostgreSQL networks, MinIO remote state, and reviewed infrastructure delivery workflows.
Persistent, searchable memory for MCP-capable AI agents, with SQLite full-text search and tools for saving, retrieving, and managing memories across sessions.
An offline-first technology learning application with interactive labs, quizzes, coding games, local learner profiles, and facilitator tools.
A Flask application that classifies uploaded audio as Akan, Dagbani, or Ikposo using mel-spectrogram features and a TensorFlow/Keras model.
A browser-based planning tool for repeatable AI agent loops, including verification gates, human review, retry policies, and JSON, YAML, or Markdown export.
A personal wellness prototype that reads Android Health Connect summaries, stores them locally, and compares daily signals with a recent baseline.
A volunteer reporting and cohort evaluation pilot with role-based access, session reports, attendance summaries, dashboards, and CSV exports.
An English speech transcription script using pretrained Wav2Vec2, audio resampling and normalisation, and optional GPU inference.
A PyTorch skin-lesion classification project with data preparation, EfficientNet-B3 training, validation reports, and experiment visualisations.
An Asante Twi speech transcription script using a pretrained Akan Whisper model with audio resampling and optional GPU inference.
A Streamlit audio classifier for Akan, Dagbani, and Ikposo, using mel-spectrogram preprocessing and a TensorFlow/Keras model.
A containerised React portfolio demonstrating multi-stage Docker builds and Nginx serving.
Infrastructure-as-code demo deploying an Nginx app with Terraform and Docker.
Provisioning a Docker Swarm cluster on Windows using Vagrant and VirtualBox.
A local PDF-to-audiobook command-line tool with chapter audio, merged outputs, and Piper or Coqui text-to-speech backends.