Ry Nduma

Building AI systems for scientific discovery.

About

I am currently an ML Intern at CuspAI, working on agentic approaches to high-throughput materials design. Previously, I worked in Aron Walsh's Materials Design Group at Imperial, where I built Crystalyse (a provenance-enforced scientific agent), co-authored a synthesis perspective in review at Nature Synthesis, and won two international hackathons including the global LLM Hackathon for Materials Science & Chemistry with Sky.

I was born and raised in Nairobi, currently in London as a final-year Beacon Leadership Scholar at Imperial College Department of Materials. In my downtime, I'm usually hiking, lifting, playing guitar, or trying to get better at Elden Ring.

Work Experience

View complete experience →

Projects

Featured

Crystalyse

A provenance-enforced scientific agent for computational materials design. Crystalyse orchestrates computational tools through natural language while ensuring every numerical result traces to explicit calculations, addressing the core challenge of AI hallucination in scientific domains.

From a terminal prompt like "Suggest a new Na-ion battery cathode", the system autonomously computes capacity (193 mAh/g) and voltage (3.7 V) in ~90 seconds—despite having no pre-coded battery workflows. The agent reasons about which fundamental calculations to chain together, then derives electrochemical properties from first principles.

The goal is to intelligently and responsibly accelerate the computational parts of materials design so scientists can focus their energy on the challenging discovery questions.

AI Agents Materials Design LLMs Scientific Computing

Sky

An agentic synthesis exploration tool using recursive search algorithms and LLM reasoning to predict synthesis routes for novel materials. Winner of the Global LLM Hackathon for Materials Science & Chemistry.

LLMs Materials Science Agentic AI

MatSX

AI-powered platform for materials similarity search and anomaly detection using Magpie composition descriptors and MACE structure embeddings. Winner of the KRICT ChemDX Hackathon.

AI Materials Informatics 3D Visualization

Materials Informatics Advanced Practical

An interactive course covering the computational materials design pipeline—from combinatorial screening with SMACT to structure generation with Chemeleon, machine learning force fields with MACE, and DFT with VASP. Built as a Jupyter Book for Imperial's Materials Department.

Education Materials Informatics Jupyter Book
View more on GitHub →

Open Source Contributions

SMACT

Core contributor. Compositional screening and property prediction for materials discovery. Added ML features for band gap and bulk modulus prediction using ROOST, compositional screening utilities, and chemical validation tools.

Open Source Materials Informatics Python

Publications

Awards & Honors

Community & Volunteer Work

Writing

I write to think clearly. thinklog is where I document ideas, learnings, and explorations in AI, Philosophy and Materials Science—long-form notes for future me, public in case they're useful to others.

Contact

This site is open 24 hours a day, 7 days a week. Thanks for visiting!

Oh! Before I go, here's a playlist you might enjoy!

LinkedIn GitHub Twitter Email