// About — Mohamed Saadi
Closing the Gap Between
Research and Production.
Most AI initiatives stall not because the research is wrong — but because the gap between a theoretical capability and a production-grade, observable, maintainable system is enormous. That gap is where I work.
My approach has never changed: find the most relevant advances from AI research, understand them at the mathematical level, then build the architecture and the practices that make them work for real enterprise teams — with real delivery constraints, real budgets, and measurable outcomes.
I've done this through every major paradigm shift of the last decade. Each time, I arrived before the market had a name for it.
"The rarest skill in AI is not knowing the research — it's knowing which research matters, and how to make it work in production, on time, for people who can't afford experiments."
01 / Proven Enterprise Impact
02 / A Pattern, Repeated Three Times
Arriving Before the Market Has a Name for It.
Growing up in an environment of rigorous scientific thinking, then through classes préparatoires and École Nationale Polytechnique d'Alger, then research at LIP6 / CNRS Paris — I developed an early instinct for identifying which research directions would matter in practice, and a systematic approach to operationalizing them.
That instinct has been validated three times, across three paradigm shifts:
Research: Distributed computing, columnar storage, stream processing
Real-time 360° Customer View platform at EDF — before Big Data was standard in French enterprise.
Research: Reproducible ML pipelines, feature stores, model governance
Full MLOps stack built from scratch on AWS & OpenShift — when the term was still a blog post.
Research: Multi-agent negotiation, neuro-symbolic reasoning, LLM orchestration
LangGraph agentic workflows, MCP, RAG at EDF Group — grounded in 2016 CNRS research on autonomous agents.
03 / The Starting Point
LIP6 / CNRS, 2016. Before It Was a Buzzword.
Growing up around medicine, my first memory of awe wasn't a stethoscope — it was a Pentium 3. That machine sparked a question that has never left me: can intelligence be engineered from first principles?
At LIP6, that question found its formal framework: Multi-Agent Systems. How do autonomous, self-interested agents — with incomplete, asymmetric information — negotiate, coordinate, and produce collective intelligence? I designed bargaining protocols and revenue-sharing mechanisms in JADE. I studied coalition formation and adaptive pricing in multi-echelon supply chains.
Today, LangGraph agents orchestrating tasks through tool calls and memory layers are solving the same coordination problem I formalized in 2016. The compute changed. The question didn't.
04 / 10 Years — Every Paradigm Shift
-
2015–16
Research — Multi-Agent Systems
LIP6 / CNRS Paris. Autonomous agents negotiating under incomplete information. Supply chain optimization. 188-page thesis. The field will name this Agentic AI a decade later.
-
2017
ML Engineering — Pharma
Sanofi. Demand forecasting for pharmaceutical distribution. 8% stock reduction in production. First experience translating ML research into regulated operational context.
-
2017–19
Wave I — Big Data
EDF. Real-time 360° Customer View platform. Apache Spark, Hadoop, Elasticsearch. Enterprise-scale data engineering before it was standard practice at French utilities.
-
2019–21
Wave II — Cloud Native & MLOps
EDF. AI/ML platform on AWS & OpenShift: 100+ daily training jobs, feature store, CI/CD, model registry. MLOps built from scratch when the term was still a conference topic.
-
2021–Now
Wave III — GenAI & Agentic Architecture
EDF. LLM integration, RAG, LangGraph agentic workflows, MCP. Governance across 10+ teams. The theoretical framework from 2016 now deployed at enterprise scale.
05 / The Technical Stack
From Algorithm to Production System.
I operate across the complete vertical — from mathematical foundations to cloud-native infrastructure. The depth at each layer is what makes the architecture decisions sound.
Agentic AI & LLMs
MLOps & LLMOps
Cloud Native & Infrastructure
API & Architecture Patterns
Governance & Engineering Leadership
Foundations
06 / Published Research
2016 · École Nationale Polytechnique × LIP6 / CNRS Paris
A Multi-Agent Negotiation Approach for Supply Chain Management
188 pages · JADE · AUML · Nash Bargaining · Gaia Methodology
The distributed negotiation protocols in this thesis — Nash bargaining under incomplete information, coalition formation, revenue-sharing — are the theoretical foundation of what production Agentic AI systems implement today under the name of "orchestration" and "tool use."
07 / What I'm Building
NeoStair EURL
Independent Consulting Practice
Senior Solution Architecture for enterprise-grade AI platforms. Current engagement: EDF DSIN Direction Commerce — GenAI platform, Agentic AI workflows, MLOps at scale.
SeriesMind
Multi-Agent Time Series Platform — early stage
Six specialized autonomous agents with human-in-the-loop validation for enterprise time series intelligence. The same 2016 research — now with LLMs, production MLOps, and a real product architecture.
08 / Beyond the Screen
The Athlete
Former professional handball player. Today, cycling — not as leisure but as commitment. Each year I ride the Tour de France routes across Europe, stage by stage. Hyrox races fill the winters.
The Mind
Iyengar Yoga — the school of absolute precision and structural alignment. Alongside sustained interests in neuroscience, cognitive science, and mindfulness practice.
The Listener
Cosmic jazz: Ibrahim Maalouf, Dhafer Youssef, Avishai Cohen. Music as the art of structured improvisation — the beauty of constraint.
The Aesthetic
Brutalism and neo-brutalist architecture. Structures that expose their raw logic and carry their weight honestly, without ornament.
Available for senior missions
Solution Architect AI · Agentic AI · MLOps · LLMOps · GenAI Platform
Paris region & remote · via NeoStair EURL
Get in touch