Loading…
Fetching repositories…
MSc Research • MLOps Portfolio
Combining academic Machine Learning research with production-grade MLOps, infrastructure automation, and rigorous testing practices.
Flagship Research & Applied ML
Audio Transcription: Wav2Vec2
Star SignalEnd-to-end ASR pipeline utilizing Wav2Vec2 for high-fidelity English audio transcription with custom fine-tuning logic.
Ghanaian NLP: Language Detection
Lightweight NLP model identifying low-resource Ghanaian dialects with high accuracy using PyTorch.
Visual CNN in Akan Twi
Dataset curation and supervised training for a localized CNN image classification model focused on Ghanaian context.
Brain Tumor Federated Learning
Decentralized model training architecture for medical imaging, ensuring data privacy across localized nodes.
Testing & Quality Engineering
✓ Unit Tests
ML Model Unit Testing
Pytest-based unit test suites for data preprocessing transforms, feature engineering functions, and model inference outputs — catching regressions before they reach CI.
⇶ Integration Tests
Pipeline Integration Testing
End-to-end integration tests for MLOps pipelines — validating data ingestion, model training, evaluation thresholds, and artifact storage in isolated environments.
▶ E2E Tests
Model Serving E2E Tests
Automated end-to-end smoke and contract tests against deployed REST endpoints — verifying latency, response schema, and model output distributions in staging.
△ Performance Tests
Load & Stress Testing
Locust and k6 load tests against inference APIs, measuring throughput and p95 latency under realistic traffic patterns, with automated alerts on regression.