Projects

Student Paper Presentations

• Pick a recent publication (or a few related ones) in the Materials + AI domain.
• Focus on topics we have not covered in class or on interesting applications.
• Sign up for a presentation slot (two available dates) on the Google Sheet (link will be provided). Sign-up is first come, first serve (you need to select a paper to sign up).
• Prof. Peter will approve each paper in the Google Sheet.
• Duration: 15 minutes (12 min presentation + 3 min Q&A).

Individual Final Project

Choose one of the two options below.

Option 1 — Full ML Pipeline

• Select a target materials class (e.g., 2D materials, perovskites, metal oxides) and a target materials property (can be experimental or computational).
• Conduct a literature review of available datasets and prior ML efforts.
• Collect relevant datasets and perform data cleaning. Matminer and Materials Project data may not be used (all other sources are ok).
• Carry out the full ML pipeline with two featurizers.

Option 2 — Atomistic Simulation with MLIPs

• Identify an atomistic simulation problem/system.
• Conduct a literature review on the appropriate modeling approach.
• Simulate the system using ML Interatomic Potentials (MLIPs).
• This can either involve molecular dynamics workflows or DFT-like property predictions involving total energy predictions.

Deliverables

Short Project Summary: 2–3 pages summarizing the relevant literature and the high-level ML pipeline (data cleaning, featurization, ML algorithm, hyperparameters, performance metrics) or MLIP approach (model chosen, summary of model predictions).

GitHub Repository:

• Contains the entire code to recreate the ML pipeline, plots, and results.
• Contains a summary JSON or CSV file with all main performance metrics.
• Contains PNGs of all relevant plots.

Final Presentation: 25 minutes (20 min presentation + 5 min Q&A).