Evidence
Product judgement, backed by inspectable work.
Three forms of proof support it: production-scale product experience, research and engineering artefacts built and tested directly, and a decision method you can inspect before you buy.
Career provenance, Hyperion-authored research and engineering, and public method are shown separately.
Connect each claim to the evidence that supports it
Buyer diligence
What the evidence lets you conclude.
Each answers a different diligence question: has he operated at scale, does he still build, and can you inspect how the decisions are made?
How to read these classes
- Production
- Observed in a live production environment carrying real operational load.
- Client-approved
- A client relationship and scope the client has approved for public description — no outcome implied unless separately approved.
- Owner-confirmed
- Relationship and bounded scope attested by Hyperion's owner. This label does not establish client approval, endorsement or an outcome.
- Pre-production
- Hyperion-owned work running end to end but not yet carrying production load — not a customer deployment.
- Controlled
- Measured under conditions chosen by Hyperion — repeatable, but not a field result.
- Simulated
- Measured against a simulated environment or workload rather than physical hardware in the field.
- Modelled
- A calculated scenario from stated assumptions. Not a measurement, not an outcome and never a case study.
- Public method
- A publicly checkable artefact — preprint, open-source contribution or published method. A preprint is not peer-reviewed.
- Career record
- Work done by the founder at a former employer, before Hyperion. Those organisations are not Hyperion clients.
Named client mandate
A named client relationship with bounded scope — no outcome or endorsement implied.
- Owner-confirmed
Schneider Electric engaged Hyperion Consulting for Software-Defined Platform work.
Founder career track record
Has operated at production scale.
- Career record
17+ years
- Career record
NDS/Cisco video platforms deployed at 100M+ scale
- Career record
Deputy General Manager, Connected Car Services at the Renault-Nissan-Mitsubishi Alliance — product leadership on connected-services programs (Renault OpenR Link, NissanConnect), technology now deployed in 7M+ connected vehicles
Hyperion-authored research references
Still builds systems directly: Auralink.
- Controlled
32 conformance cases
ACS v0.1 covers reference, safety and federated-operation scenarios. The suite is engineering evidence, not certification or a field result.
Inspect the working evidence
The method is inspectable before an engagement.
Inspect the supporting claim record
Founder career track record
Founder experience at former employers — not Hyperion clients.
- product marketing at ABB E-mobility on EV-charging hardware and charging software
- MIT Professional Education — eight professional course certificates (2016–2017), including Data Science, Systems Engineering, Cybersecurity, and IoT
- Worked on embedded Linux, Android and open-source integration for production connected-device platforms during his NDS/Cisco career.
Hyperion-authored research references
Authored and documented by Hyperion to make the method inspectable — not external client deployments.
- in-house AI ventures, including Auralink
External references
Third-party published analysis — cited, not a Hyperion measurement.
- the majority of AI pilots stall before production
Every engagement is led directly by Mohammed Cherifi. Specialist partners may be introduced for clearly defined work when required and disclosed; Hyperion is not a staffing or engineering-delivery bench.
Selected engagement
One client, described exactly.
Consultant & Tech Lead, Software-Defined Technologies — Schneider Electric, Innovation & Technology. January to April 2024. Developing a strategic framework for software-defined technology capability; leading and coordinating innovation squads through delivery; introducing agile working methods into the software development process; and structuring a new software activity against business requirements.
This describes the mandate and the role. It does not disclose information shared in confidence, claim a client outcome, or constitute a public case study.
How we describe our work
Every organisation named on this site falls into one of five classes. The class is stated wherever the name appears, so no reader has to infer the relationship.
- Verified public client
- A client relationship Hyperion may state publicly: the engagement is documented and the governing contract places no restriction on reference. Described as mandate and role — never as an outcome. Where a client has given written permission, the evidence register records it.
- Verified confidential client
- A real engagement that may not be named. Described only in the abstract, if at all.
- Partner or subcontractor
- Work delivered with or through another party. Never presented as a Hyperion client.
- Prospect or lead
- A conversation, proposal or target. Never presented as delivered work.
- Founder career experience
- Organisations Mohammed worked for as an employee. Not Hyperion clients, and labelled as career wherever their marks appear.
No client is named on this site unless the engagement is documented and the governing contract places no restriction on reference. Displaying an organisation's logo is treated as a separate question from naming it, and is not implied by any class above.
Relationships, stated precisely
Career provenance, Hyperion-authored research and engineering, and public method are shown separately.
French Government AI Ambassador for Industry — Osez l'IA Initiative
Berkeley SkyDeck Key Advisor (pro-bono, since October 2022)
FranceNum Activateur
Board Member, La French Tech Athens
Also built by Hyperion
Other Hyperion-built systems
A plain list, kept outside the classified register above. Each line states the system's stage. None is client evidence, and no vendor partnership is implied.
Described on the Ventures page
Other systems
Auralink OS
A proposed open operating system for Physical AI and Industrial AI machines, in which AI proposes and a small deterministic authority on the machine decides.
Design phase; nothing is assessed or certified.
Hyperion Physical AI Workbench
A local-first tool that profiles edge hosts, catalogues models, selects runtimes, and deploys and rolls back advisory Physical AI workloads.
Local-first prototype; not a real-time or safety controller.
Hyperion Data Factory
A proposed local-first data factory for Physical AI and Industrial AI: physical scenarios, simulated sensors, injected faults and traceable datasets.
Research and design; no generator is implemented yet.
GUIDE
Helps EU Physical AI and Industrial AI organisations frame their next AI-transformation decision: what to decide, what evidence is missing, and who should examine it.
Prototype in private development; not live.
Pharodos
A Greek-first diagnosis tool that turns a small business's self-reported need into an explained next step, simpler alternatives and a seven-day observation plan.
Prototype; not launched.
Ereisma
A Greek legal-assistant prototype for statute research, citations and draft templates, built and tested on synthetic data only.
Prototype; generated documents require lawyer review; no accuracy claim is made.
Reachy Mini voice bridge
Speech in, spoken answers and expressive head and antenna movement out, for the Reachy Mini desk robot.
Prototype; on-robot validation is pending. Reachy Mini is a Pollen Robotics product; no partnership is implied.
Open-weight skills marketplace
Agent skills, an SDK and an MCP server that run on open-weight models served locally or through an API.
Prototype.
Open the marketplaceHyperion Engineering System
How Hyperion builds: several AI models draft and challenge each other's work, while deterministic tests, linters and review gates keep final authority and evidence records.
Internal practice.
Judge it against your product
The fastest way to evaluate the judgement is to put one of your own product decisions in front of it.