ML and lab automation, in public

Roadmap

1. nowFair model comparison on a small datasetRidge vs. random forest vs. neural net, with a cross-validation scheme that suits small experimental data
2. nextUncertainty and interpretabilityHow sure is the model, and which inputs does it rely on?
3. laterEnd-to-end Raman spectra pipelineIngestion, preprocessing, training, evaluation and experiment tracking with reproducible configs
4. laterActive learning, simulated offlineLet a model pick the next experiment, tested on public or synthetic data first
5. laterInstrument control and a closed loopPython runs the measurement; the model suggests, the instrument runs, the model updates

TODO(Sidney): confirm the order and which step is “now”.