Rayen
Lassoued

AI/ML Engineer & Full-Stack Developer — Multi-Agent Systems, MLOps & Backend Engineering

I build systems that reason in fleets — autonomous agents, the pipelines that keep them honest, and the full-stack interfaces that make their work legible.

Teaching a
consultancy to
write its own
proposals

Development-agency tenders are won on precision under deadline. At Madiba Consult I built the machinery behind GIZ proposal production: an orchestrated set of LLM agents that reads a terms-of-reference document, decomposes it into a compliance matrix, drafts each section against the house voice, and hands a reviewer a traceable document rather than a plausible one.

The second half of the problem was people. Expert matching runs as retrieval over a corpus of roughly twelve thousand consultant CVs — embedded, filtered against mandatory qualifications, then re-ranked so a proposal manager sees six defensible candidates instead of six hundred results. Every recommendation carries its evidence back to the source line.

Underneath both systems sat a document-processing pipeline I built to make the raw inputs usable: parsing and structuring hundreds of tender PDFs and consultant CVs into clean, section-aware text before anything ever reached an agent.

Each entry is a system I designed, built, and put in front of users or in production. Filter by domain.