Patent Alchemy

Inspiration

We got fired up because two papers made the problem obvious. One from the Business and Sustainable Development Commission talked about a massive market opportunity around sustainable solutions, and another supply chain white paper from MPO showed that most products still move without real end to end visibility. Those two facts together felt like a dare - there is huge value sitting in patents and no good way to turn it into real, usable product data.

What it does

It turns a patent from a dusty PDF into a living product passport you can actually use. The app pulls out the useful design pieces, maps them to materials and variants, runs a quick carbon estimate using published sector factors, and builds a structured passport you can store or resolve. The idea is simple and useful: translate patent prose into a BOM, show the CO2 numbers, and package it all as a shareable, auditable record.

How we built it

We built a web workbench with a Python backend and a modern frontend. The backend parses patent text and runs smart models and heuristics to split patents into functional blocks. We match blocks to material options, use public LCA factors for carbon screening, and emit a JSON digital passport with GS1 style identifiers and an events array for traceability. Frontend is made so you can load a sample, pick materials, run screening, and mint a passport all on your laptop.

Challenges we ran into

Patents are messy and weird. The prose is legal and full of edge cases, PDFs come in strange formats, and some glyphs or fonts break text extraction. Mapping vague language in claims to actual materials is hard and needs careful heuristics. Getting trustworthy carbon factors and citing methods in a transparent way took work. Making the demo look slick also meant fighting rendering and tooling problems when producing animations and the intro video.

Accomplishments that we're proud of

We built a full end to end loop you can run live on a laptop. From loading a patent to extracting blocks to producing a BOM and a CO2 screening, and finally minting a structured passport, everything is there. The passport includes provenance back to the patent and a clear method note for the footprint. The UI is interactive enough for judges to try a what if in 30 seconds and see real numbers change.

What we learned

Clear provenance matters more than we thought. If you show a number without the method, people will distrust it. Small UI decisions massively change how fast someone trusts and explores the tool. Also, patent text needs lots of fallbacks and conservative parsing so you don’t hallucinate materials. Finally, integrations like GS1 and EPCIS are doable but require attention to standards early on.

What's next for Patent Alchemy

Make extraction way more accurate, expand the material and factor database, and plug into real supply chain feeds so passports can update automatically. We want to add enterprise export hooks and an easy resolver so a passport can be fetched by any phone or ERP. Long term we want this to be the fastest way for a product team to go from an idea buried in a patent to a traceable, sustainability-aware product spec.

Built With

• fastapi
• github
• iai
• nextjs
• petlca
• uvicorn
• wipo