AI scoring and editing for technical texts
Focus area
AI & automationClient
Under NDA
Timeline
1 year
Tools
Python, FastAPI, LLM + RAG, pgvector, React
60%
more grants won
>5,000
sources analysed
>8
research groups using the tool secured funding
9.3/10
ease-of-use score from staff
The challenge
Applications were reviewed by hand, the outcome depended on who reviewed them, and authors had no idea how to make their case stronger.
Solution
We studied more than a thousand applications and built a scoring system plus an assistant that checks and edits the text. Both are part of the organisation’s workflow.
Result
Scoring became more accurate than a leading foreign alternative. Staff rated the system 9.3/10 for ease of use, and every application now lives in one place.
Next project
Investor PLG: apartment returns in one app
Digital products