Hands-on Tutorials

13 coding sessions on Google Colab — from molecular fingerprints and GNNs to Bayesian optimization and chemistry agents. Click any card to open the tutorial.

Week 1 · Aug 10 – 14 · Foundations of ML & Representations

Molecular Property Prediction

Building a pipeline using RDKit, feature engineering, training and performance evaluation

Open in Colab →

Model Training & Performance Evaluation

Train/val/test splitting, data leakage, chemically-relevant splits, baselines

Open in Colab →

Feed-Forward Net — Formation Energy

PyTorch from scratch, training loop, early stopping on materials data

Open in Colab →

Molecular Property Prediction with GNNs

PyG on QM9, MPNN for HOMO-LUMO gap, atom embedding UMAP

Open in Colab →

Explaining ML Predictions

SHAP values on molecular fingerprints, attention visualization on GNNs

Open in Colab →

Few-Shot Learning

Transfer learning, application, failure modes of limited-data regimes

Open in Colab →

Molecular Generation with VAEs

Encode/decode molecules, latent space exploration, validity checks

Open in Colab →

Week 2 · Aug 17 – 21 · Dataset Design & LLMs

Fine-tuning ChemBERTa

HuggingFace, BACE IC50, frozen vs. full fine-tuning comparison

Open in Colab →

Active Learning for DFT Triage

QM9/ANI-1 pool, GP uncertainty sampling, learning curves

Open in Colab →

Reaction Yield Optimization with BO

Suzuki coupling dataset, BoTorch, EI vs. UCB convergence comparison

Open in Colab →

Uncertainty Quantification in Practice

Calibration curves, conformal prediction on molecular datasets, ensemble variance

Open in Colab →

Chemistry RAG Assistant

Index papers, Q&A on reactions, RDKit tool calls

Open in Colab →

Building a Chemistry ReAct Agent

PubChem lookup, RDKit tools, BO surrogate, guardrails

Open in Colab →