/ Works
/ 2025
/ LMA EU Hackathon
VERDEX
An AI-powered platform validating transition finance projects against LMA frameworks, bridging African project developers with global climate capital.
/ Client
LMA EU Hackathon
/ Role
Full Stack Developer
/ Timeline
Dec 2025 — Jan 2026
/ Stack
LMA EU Hackathon / Next.js / Pinecone / Claude / Python / TailwindCSS / Vercel / Railway

About the project
/ Overview
VERDEX is an AI-powered validation platform that helps African project developers align their transition finance proposals with international LMA (Loan Market Association) Green Loan Principles. It uses a RAG pipeline to analyze project documents against regulatory frameworks and provides actionable compliance feedback.
/ The Challenge
Transition finance in Africa faces a massive gap — project developers struggle to meet international compliance standards, and investors lack tools to validate projects at scale. The challenge was building an AI system that could accurately interpret both technical project documents and complex financial regulatory frameworks, then provide reliable compliance assessments.
/ The Approach
I built a RAG (Retrieval-Augmented Generation) pipeline using Pinecone for vector storage of LMA framework documents and Claude AI for intelligent analysis. The frontend is a Next.js app with TailwindCSS, deployed on Vercel. The Python backend on Railway handles document processing, embedding generation, and orchestrates the AI validation workflow. Users upload project proposals and receive detailed compliance reports with specific recommendations.
/ The Result
Won recognition at the LMA EU hackathon. The platform successfully validates project proposals against LMA frameworks with high accuracy, reducing the compliance assessment process from weeks of manual review to minutes. Currently live and processing real project submissions.
/ What it took
- RAG pipeline with Pinecone vector search for LMA framework document retrieval
- Claude AI integration for intelligent compliance analysis and recommendation generation
- Document processing pipeline — PDF parsing, chunking, and embedding generation
- Next.js frontend with real-time validation progress and detailed compliance reports
- Python backend on Railway orchestrating the multi-step AI validation workflow
/ Results and outcomes
- LMA docs referenced
- 6,000+
- Vector DB indexed
- 20+
- Not weeks of review
- Minutes