Accretion is how the universe builds.
Dust and matter spiral together under gravity until planets and stars form. We named the company after that process because it is our thesis in one word: individual atoms, assembled by intelligence, into new materials.
Every material starts as a search
The space of possible compounds is larger than the number of atoms in the galaxy. For two centuries, humanity has explored it by hand: mix, heat, measure, repeat. Most of the map is still dark.
Intelligence is the force that pulls
Models trained on physics — simulation data, published syntheses, measured properties — can feel the shape of that dark space. They pull promising structures out of combinatorial dust the way gravity pulls planets out of a disk.
The outcome is a material you can hold
A ranked candidate, a synthesis route, a validated sample. Computation compresses the decade between idea and material into days. That compression is the company.
Karthikeya Aduru
FOUNDER & CEOKarthikeya Aduru is the founder and CEO of Accretion Labs, a deep-tech startup building AI-assisted tools for materials discovery. A student founder, he started researching graph neural networks and materials science in 2024 before incorporating the company in 2026. Accretion Labs is DPIIT-recognised and currently building its first proof of concept.
LinkedIn ↗Claims tied to mechanism, never to adjectives.
Computation earns trust only when the lab confirms it.
Built in India. Built for everywhere matter matters.
A young field, moving fast, with room for more than one winner.
Google DeepMind's GNoME and Microsoft Research's MatterGen and MatterSim proved the model-layer approach at lab scale. A newer cohort of specialists is now applying the same toolkit to narrower, high-stakes verticals:
Schrödinger
The longest-established name in the space, extending a physics-based simulation platform with machine learning across both materials and drug discovery — a reminder that the ML layer works best paired with a physics engine that already understands the domain.
Orbital Materials
An AI-native startup applying generative design and simulation to clean-energy and environmentally sustainable materials — a narrower, deeper bet than general-purpose discovery.
CuspAI
Another AI-native entrant applying the same generative-plus-simulation pattern to carbon-capture materials — a vertical-specific bet on the same underlying toolkit.
We read that as validation, not competition. The techniques are converging into a shared toolkit; what differs is which problems a team chooses to point them at, and how carefully the lab checks the model's homework. We built Accretion around the second part.
We're assembling a small team of computational and materials scientists.
If you work at the intersection of ML and physical chemistry — or want to — say hello.