Skip to content

Netherlands · Delft

Physical AI Product Leadership from Delft

Hyperion helps Dutch robotics, industrial-automation, mobility and energy teams turn technically promising systems into dependable products.

Working languages: English and French.

Product and technology ownership gaps Hyperion resolves

Each of these is a decision nobody currently owns end to end. Engineering can keep progressing while the decision stays open.

  • The demo works. The product is undefined.

    A cell, a vehicle or a line performs under chosen conditions. Nobody has written what the complete product must do for the customer who has to buy, operate, support and trust it.

  • Hidden work is holding up the business case.

    Expert tuning, operator intervention, bespoke integration and unpriced support keep the pilot alive and never appear in the unit economics.

  • Architecture is committed before intended use is settled.

    Field boundaries, service model, safety envelope and update strategy surface after the expensive commitments are already made.

  • No named owner across model, machine, operator and business.

    Four functions each own a part. The trade-off between them belongs to nobody, so it is resolved late and by default.

The Deployment Gap: three Dutch market wedges

The distance between a working pilot and a repeatable product is where Dutch physical-AI value is won or lost. Three places it is most expensive:

  • Grid-constrained charging & industrial electrification

    Congestion has made flexibility, orchestration and capacity product problems. Charging and electrification companies are becoming software-and-services businesses mid-flight.

  • Intralogistics & intelligent equipment going software-defined

    One reference site is not a fleet. Product-family boundaries, brownfield integration and recurring software packaging decide whether equipment scales as a product or stays a project.

  • Asset-heavy operators deploying Physical AI

    Ports, logistics and industrial operators face vendor due diligence, pilot governance and the scale decision — with no internal owner for the complete product question.

Read the full market thesis

The full thesis — definition, five problematics, market evidence with sources, and limitations — is published ungated in English.

Product Decision Review

A bounded review that settles one consequential product decision with evidence. Two to three weeks. You receive an executive decision memorandum, an evidence and risk map across the six lenses, and a prioritised 90-day plan: go, redirect, sequence or stop.

Explore the Product Decision Review

Fractional CPO / Interim Head of Product

Embedded product and portfolio leadership when you need a senior owner now but not yet a permanent executive. Product strategy, portfolio, roadmap, customer and operator value, product economics, prioritisation, productization and go-to-market — inside an explicit executive mandate with a written stop condition and handover owner.

Explore the Executive Product Leadership

AI transformation and selected Physical AI CTO scope

Where the ownership gap sits beside product rather than inside it, a mandate may extend to AI transformation / Fractional Chief AI Officer scope — AI portfolio priorities, operating model, build/buy/partner decisions, evaluation, governance and measurable scale-or-stop decisions — or to Interim CTO responsibility for Physical AI architecture, engineering direction, reliability, security and production readiness. The title follows the ownership gap; the method remains one integrated product system.

Dutch priority sectors

Where Physical AI product decisions are consequential and the field evidence is hard to get.

  • Delft · RoboHouse · RoboValley

    Robotics teams moving from a working prototype to a product an operator can run without the research group in the room.

  • Rotterdam port, maritime and logistics

    Autonomy, perception and terminal automation, where degraded conditions and safety cases decide what can be deployed.

  • Brainport high-tech systems and manufacturing

    Intelligent equipment and high-mix production, where product architecture and serviceability decide unit economics.

  • EV charging, energy and connected mobility

    Software-defined charging and grid-edge systems — the domain of the founder's ABB E-mobility career and of the public preprint.

  • Westland horticultural automation

    Vision, manipulation and autonomy in variable biological environments, where the acceptance bar is operational, not benchmark.

  • Semiconductor and advanced equipment

    Precision systems where an AI-enabled subsystem must meet an existing reliability and qualification regime.

  • Amsterdam AI scale-ups

    Teams whose model works and whose product, pricing and operating boundary are still open questions.

  • Utrecht product and technology market

    Product and platform organisations introducing AI inside explicit product, update and safety boundaries.

Evidence

Each item is labelled for what it is. Client work, founder career, public appointment, public research and internal R&D are different classes of evidence and are never blended.

Client mandate
One bounded client mandate on Software-Defined Platform work, January–April 2024, as Tech Leader within Innovation & Technology, Energy Management. The named client disclosure is published in one canonical place — the evidence page — and no case study, quotation, result metric or endorsement exists for it.
Founder career
17+ years at NDS/Cisco, Renault-Nissan-Mitsubishi and ABB E-mobility. Roles held at former employers — not Hyperion client engagements.
Method
The Hyperion Physical AI Product System — five decision gates and six evidence lenses governing when a product may advance.
Public research
arXiv:2603.08736, a public technical preprint on autonomous edge-deployed AI agents for EV charging infrastructure. A preprint that has not been through journal peer review; its results are a controlled-environment evaluation.
Internal R&D
Auralink — Hyperion-owned pre-production R&D awaiting hardware integration. Not a customer deployment and not commercial outcome evidence.
Internal R&D
A Reachy Mini / SO-101 imitation-learning bench. The perceive–understand–interpret pipeline is measured on the author's own real-scene captures; physical on-robot execution is not measured, and command dispatch is validated in dry-run only.

Independent mandate structure

Every engagement is written down before it starts.

  • One scope, and the decision it exists to settle
  • Explicit decision rights and an executive sponsor
  • A named evidence gate and acceptance criteria
  • A stop condition either side may invoke
  • A named handover owner and an exit condition
  • No engagement requires a follow-on engagement

Hyperion is an independent practice registered in France and operated from Delft. It does not resell hardware or software, takes no vendor commission, and does not provide legal advice, certification or conformity assessment.

Bring the decision you cannot leave unowned.

Describe the product, the blocked decision, its deadline, the executive sponsor and the evidence currently available. Mohammed will tell you whether Hyperion is the right mandate—and when it is not.