·952 words·5 mins
Open-source Python serialisation library for scikit-learn that decomposes fitted estimators into JSON and safetensors.
·1631 words·8 mins
Streamlit app that detects faces with RetinaFace and sticks googly eyes on them, using ONNX Runtime for fast inference.
·896 words·5 mins
PyTorch VAE that learns a person’s handwriting from a few strokes and synthesises new characters in their style.
·664 words·4 mins
XGBoost model that predicts specialty coffee ratings from roaster, origin and flavour notes, deployed on AWS Lambda.