About
Built so that every number can be defended
FoodProt AI helps a research team turn published papers into a dataset they can analyse and cite, with the evidence for each value kept beside it.
Why it exists
Extracting data by hand cannot keep up with the literature, and fully automatic extraction loses track of where a value came from. The same property is reported under different names, units and conditions, so two numbers are comparable only when their context is kept with them.
FoodProt AI does the repetitive part — searching, collecting, reading, standardising — and leaves every scientific decision with a person.
What it is not
- It does not replace your review protocol. It runs it and records it.
- It does not judge study quality. It records what a paper reports, and where.
- It does not promise that every paper can be collected or every value extracted.
- It does not get around publisher subscriptions or library logins.
Principles
- A suggestion is not a result
- Automated steps propose. A named person approves before a value enters a dataset.
- The record is made as the work happens
- Searches, decisions, sources and reviews are written down at the time, not reconstructed later.
- Flag rather than guess
- An unclear unit, a missing condition or conflicting evidence sends the value to a person.
- Private by default
- A project is visible only to the people its owner invites.
FoodProt AI and the FoodProt database
They are two separate products. FoodProt is a database of protein functionality data; FoodProt AI is the workspace where a team builds a verified dataset from the literature. They have separate accounts and share no sign-in or data. A dataset moves from one to the other only when a person exports it.