ai-grant-researcher·Natural Language Processing
AI Grant Researcher
Natural Language Processing Development
A multi-agent AI system built on Claude for automated preparation of grant applications to European programmes — from parsing the call to a ready submission package.
Details
General information
- Direction
- Natural Language Processing
- Stage
- Development — Building and iteratively improving the solution.
- Technologies & methods
- Multi-agent AI, Anthropic Claude, TypeScript / Node.js, Prisma / PostgreSQL, React-PDF
- Format
- Research and development together with students and the Center's laboratory.
Overview
About the project
The system runs the full application cycle through five specialized agents: Grant Analysis (parsing the call, extracting requirements and evaluation criteria), Feasibility (assessing the organization's real chances and eligibility, including correct handling of Ukraine's status as an associated country), Scientific Writing (generating the scientific part), Reviewer (expert assessment and scoring against EU criteria), and Application Package (producing administrative documents). An iterative "write → review → revise" pipeline runs until a quality threshold is met; it supports direct import of call data from the EU Funding & Tenders Portal (SEDIA API), auto-filling of official EU templates (Part A / Part B), and PDF generation. It covers Horizon Europe, the Digital Europe Programme, EuroHPC, EIT HEI, and Erasmus+, and works in three languages. Stack — TypeScript / Node.js (pnpm monorepo), Prisma + PostgreSQL, Anthropic Claude (Opus / Sonnet), React-PDF, Python. The system has been adopted into ZNU's grant-application workflow — applications for Horizon CL4, EIT HEI, ROB4GREEN, FF+, COP-PILOT, and other calls have been prepared with it.
Stack
Technologies & methods
Multi-agent AIAnthropic ClaudeTypeScript / Node.jsPrisma / PostgreSQLReact-PDF
Interested in this direction?
Let's discuss collaboration, joint research, or a topic for coursework.