Our vision
The world runs on polymers. We are building the AI to reinvent how we discover them.
A new polymer still takes a decade to discover and qualify. RePolymer is an AI research platform for polymer science, designed to bring property prediction, inverse design, and molecular reasoning into one evidence-grounded system, so every answer comes with the structure and data behind it.
By the numbers
- 460Mt/yr
Global plastic production, and rising
- 10–20yr
Lab to market for a new material
- ~9%
Share of all plastic waste ever recycled
The scale problem
460 million tonnes a year, and the chemistry has not kept up
0M+
tonnes produced every year; only the bright sliver is ever recycled
~0%
of all that plastic is ever recycled, the rest is landfilled, burned, or lost
Every year, humanity produces roughly its own combined weight in new plastic, yet only a fraction is ever recycled, and waste systems are falling further behind production growth. The bottleneck is not ambition. Polymer discovery still moves on 10 to 20 year timelines, while the materials we depend on grow more complex, more mixed, and harder to recover.
This is a generational problem. We need a fundamentally faster way to understand, design, and validate polymers at the molecular level, before the next wave of materials is locked into another century of waste.
Sources: OECD Global Plastics Outlook (2022) · Geyer et al., Science Advances (2017) · National Academies, MGI report (2023)
The acceleration
Discovery on a 10–20 year clock. Built to compress it to months.
Traditional polymer development crawls through years of synthesis, characterization, and iteration. RePolymer is designed to collapse that loop, reasoning across structure, properties, and evidence at once, so validated candidates can arrive in months, not decades.
Sources: US Materials Genome Initiative · Audus & de Pablo, ACS Macro Letters (2017) · National Academies, MGI report (2023)
The approach
Built like a lab. Not like a chat app.
Most AI-for-science tools stop at fluent text. RePolymer starts where real polymer work does: repeat-unit chemistry, processing windows, property tradeoffs, and end-of-life pathways. Property prediction, cheminformatics, knowledge-graph retrieval, and domain reasoning are designed to work as one system, so every answer stays tied to structure, mechanism, and evidence.
We are building the infrastructure polymer informatics has lacked: standardized representations, property models validated against measured data, and workflows scientists can trust.
The stack
From polymer structure to manufacturing decision
A polymer is not a molecule on a page. It is a hierarchical, process-dependent system whose properties emerge across scales, which is exactly why generic LLMs fail in polymer science. RePolymer is built to encode backbone chemistry, side-chain contributions, formulation context, and lifecycle constraints into one representation the model can reason over.
Screen polymer candidates, troubleshoot a batch, evaluate additives, plan recovery routes: the platform is built to reason across the full polymer lifecycle in one system, with no handoffs between disconnected tools.
Polymers are step one, not the destination. Mapping a repeat unit to how it behaves is structure-to-property reasoning, and that generalizes to any material whose function is set by structure. We start where the need is most acute, with polymers and biopolymers, then build outward toward an intelligence layer for materials science, and the many industries still running on trial and error.
Why we build
Serious science. Built for the scientists who live it.
Scientists have lived the pain of slow iteration, scattered literature, and tools that do not speak their chemistry. RePolymer is built for them, because the world cannot wait another twenty years for each breakthrough: a recyclable polymer, a safer coating, a biomaterial that breaks down when it should.
This is the platform polymer science has been missing: rigorous enough for the lab and fast enough for industry.
Join us
Building this with us?
If this is the future of polymer science you want to see, we would love to build it with you.