# Ouahabi Benhenni

> Ouahabi Benhenni is an AI transformation consultant, AI architect, backend architect and technical lead (AI & backend). He helps companies, startups and institutions go from diagnostic to production: AI strategy and roadmaps, multi-agent systems, RAG for regulated environments, genomic AI pipelines, real-time backends and multi-tenant SaaS. Based in Algeria, works in Algeria, France, Europe, North Africa and remotely, in Arabic, French and English.

**Open to consulting engagements.** Contact: contact@ouahabi-benhenni.com

## Engagement formats
- AI diagnostic: feasibility audit — where AI creates value, with which data, cost and risks.
- Scoping & roadmap: requirements, use-case prioritisation, AI transformation roadmap per domain.
- Architecture design: AI architecture (agents, RAG, ML pipelines) and backend architecture (APIs, microservices, real-time, multi-tenant), with documented decisions.
- Technical lead: leading a team from design to production.
- Team training: applied AI, agents, backend architecture.

## Roles and track record
- AI Consultant · AI Project & Product Lead · Systems Architect at an AI consulting and engineering firm since August 2025: about ten AI projects led from needs analysis to production; teams of up to 15 people; sectors: pharma, industry, documentation, sales intelligence, education.
- AI Research Engineer, Genomic Medicine & Precision Oncology at a biomedical research institute (2026): pipeline from patient genomic data to LLM-generated medical report.
- Software engineer for a public administration (mail management, corruption risk mapping), full-stack developer on a multi-tenant B2B SaaS, freelance AI & full-stack developer (2023–2025).

## Representative systems
- Multi-agent OCR for multilingual (Arabic/French/English) document extraction, with deterministic validation.
- Pharma Copilot: AI strategy and roadmap (R&D, quality, regulatory, data); GxP-compliant RAG assistant scoping (versioned sources, citations, audit log).
- Dataset Intelligence Engine: multi-agent dataset analysis, structuring, labelling and documentation.
- Adaptive learning / AI-powered LMS: Python microservices, ONNX, REST, integrated into an existing Flask/Next.js platform.
- Real-time multi-room voice platform: Socket.io + WebRTC/mediasoup, voice RBAC; 218K requests in 3 hours on 2 vCPU / 2 GB RAM.
- Multi-tenant B2B SaaS: Django 5, DRF, Channels, Celery, JWT, RBAC.

## How he thinks
- The problem before the model; sometimes the answer is not AI.
- Constraints (budget, compliance, hardware) are the specification.
- AI inside the system: narrowly scoped agents, verifiable outputs, deterministic stages around them.
- Human in the loop in healthcare and regulated settings; every answer traceable.
- The simplest architecture that holds in production; decisions written down with their reasons.
- Teaching is part of delivery.

## Research
- First author: "An End-to-End Approach for SNV Pathogenicity Prediction via Progressive Fine-Tuning of Nucleotide Transformer-v2" (national conference).
- 1st place, National AI Competition 2025 (medical computer vision).

## Expertise pages
- [AI transformation](https://ouahabi-benhenni.com/en/ai-transformation/): AI transformation consultant: diagnostic, use cases, roadmap and support all the way to production. For companies, startups and institutions in Algeria, Europe and remotely.
- [AI architecture](https://ouahabi-benhenni.com/en/ai-architect/): AI architect: designing multi-agent systems (LangGraph), RAG assistants for regulated environments, and ML and LLM pipelines that are reliable and traceable in production.
- [Backend architecture](https://ouahabi-benhenni.com/en/backend-architecture/): Backend architect and backend lead: designing APIs, microservices, real-time systems (WebRTC, WebSockets) and multi-tenant SaaS platforms that perform, even under tight hardware constraints.
- [Technical lead](https://ouahabi-benhenni.com/en/technical-lead/): AI and backend technical lead: leading teams of up to 15 people, work breakdown, mentoring, review and production rollout. AI and full-stack developer when hands-on work is needed.
- [Transformation IA](https://ouahabi-benhenni.com/transformation-ia/): Consultant en transformation IA : diagnostic, cas d'usage, feuille de route et accompagnement jusqu'à la production. Pour entreprises, startups et institutions, en Algérie, en Europe et à distance.
- [Architecture IA](https://ouahabi-benhenni.com/architecte-ia/): Architecte IA : conception de systèmes multi-agents (LangGraph), d'assistants RAG en environnement réglementé, de pipelines ML et LLM fiables et traçables en production.
- [Architecture backend](https://ouahabi-benhenni.com/architecture-backend/): Architecte et lead backend : conception d'API, microservices, systèmes temps réel (WebRTC, WebSockets) et plateformes SaaS multi-tenant performantes, même sous forte contrainte matérielle.
- [Lead technique](https://ouahabi-benhenni.com/lead-technique/): Lead technique IA et backend : conduite d'équipes jusqu'à 15 personnes, découpage, mentorat, revue et mise en production. Développeur IA et full-stack quand il faut mettre la main au code.

## Articles
- [Handling 218,000 requests on 2 vCPUs: backend architecture under constraints](https://ouahabi-benhenni.com/en/articles/architecture-under-constraints/): Field notes on a real-time voice platform (Socket.io, WebRTC, mediasoup) that handled 218,000 requests in 3 hours on a 2 vCPU / 2 GB RAM server: the architecture decisions that matter.
- [From proof of concept to production: what really changes in an AI project](https://ouahabi-benhenni.com/en/articles/from-poc-to-production/): Why so many AI projects stop at the demo, and how to scope, architect and hand them over so they run in production. Field notes from an AI transformation consultant.
- [RAG in regulated environments: compliance is an architecture requirement](https://ouahabi-benhenni.com/en/articles/rag-in-regulated-environments/): How to design a RAG assistant acceptable in a regulated environment (pharma, GxP): controlled corpus, permission-filtered retrieval, sourced answers, audit log and a human in the loop.
- [Tenir 218 000 requêtes sur 2 vCPU : l'architecture backend sous contrainte](https://ouahabi-benhenni.com/articles/architecture-sous-contrainte/): Retour d'expérience sur une plateforme temps réel avec voix (Socket.io, WebRTC, mediasoup) qui a traité 218 000 requêtes en 3 heures sur un serveur de 2 vCPU et 2 Go de RAM : les décisions d'architecture qui comptent.
- [De la preuve de concept à la production : ce qui change vraiment dans un projet d'IA](https://ouahabi-benhenni.com/articles/de-la-poc-a-la-production/): Pourquoi tant de projets d'IA s'arrêtent à la démo, et comment les cadrer, les architecturer et les transmettre pour qu'ils tournent en production. Retour d'expérience d'un consultant en transformation IA.
- [RAG en environnement réglementé : la conformité est une exigence d'architecture](https://ouahabi-benhenni.com/articles/rag-environnement-reglemente/): Comment concevoir un assistant RAG acceptable dans un environnement réglementé (pharma, GxP) : corpus maîtrisé, recherche filtrée par droits, réponses sourcées, journal d'audit et humain dans la boucle.

## Profile
- [Profile (English)](https://ouahabi-benhenni.com/en/): full profile, systems and way of thinking
- [Profil (français)](https://ouahabi-benhenni.com/): profil complet, systèmes et manière de penser

## Elsewhere
- [LinkedIn](https://www.linkedin.com/in/ouahabi-benhenni-a28bb5155)
- [GitHub](https://github.com/Greyma)