Ferre Torres

A hydrogel network — cross-linked fibres and glowing nodes in violet and amber on a dark ground

We help companies integrate AI.

Experts in AI integration and AI-first applications — strategy, assistants and agents, and software built into how your company already works.

Ferre Torres BV is a Netherlands-based AI consultancy — experts in AI integration and AI-first applications. We help companies put AI inside how they already work: strategy, assistants and agents, a private knowledge core, and finance intelligence, from first workshop to production.

Recent work includes an assistant platform in daily enterprise use, a private knowledge core over company documents, and a finance platform whose chat explains forecasts and deviations. We also build physics-based digital twins when a system calls for them — one more capability the team is experienced in.

Delivered in: pharma · energy · mobility · finance · legal · industrial R&D


01 / Services

Services

AI integration & consultancy

  • Strategy, workflow selection, and architecture — from the first workshop to production.
  • Start with one proof point, then scale it into infrastructure your team owns.
  • Senior, founder-led delivery: discovery, build, evaluation, and handover.
  • AI that fits the tools, data, and controls your company already runs on.
  • Training for teams — how AI systems work, the protocols to run them safely, and how to adopt them well.

Assistants, agents & knowledge

  • Company-wide assistants with retrieval, search, meetings, and tool use.
  • A private knowledge core over your documents, metrics, and meetings.
  • Agents that act inside explicit tool, approval, and audit boundaries.
  • Versioned prompts, evaluations, guardrails, and cost controls — behavior measured, not assumed.

AI-first applications

  • Finance intelligence where chat explains the deviation — what moved, why, and what needs action.
  • Internal tools and productized workflows that turn expert work into software.
  • Discovery platforms — agentic search for recyclable materials under regulatory constraints, with human review.
  • Document intelligence that turns contracts, protocols, and filings into decisions.

Physics AI & digital twins

  • One capability among the rest: when a system is governed by physics, we build the model that way.
  • Real-time surrogates that stand in for slow simulations — fast enough to explore and interact with.
  • Digital twins that track a system from its data and correct themselves when the model drifts from reality.
  • The physics stays inside the model, so its predictions stay realistic even far from the data it learned on.

02 / Work

Selected work

A de-identified finance analytics dashboard: a revenue trend chart, metric tiles, and an AI insights panel
Fig. 01 — Finance intelligence platform

Finance intelligence platform

System
An executive finance cockpit for a European group — revenue, orders, cash, and alerts in one view.
What it does
An AI panel explains each deviation: what moved, why, and what needs action — grounded in the company's own numbers.
Outcome
Board-ready commentary and answers in seconds, without exporting to a separate BI tool.

From our client work — rebuilt with no client data, numbers, or branding.

A de-identified enterprise AI assistant: chat history sidebar, model and sources controls, and quick actions
Fig. 02 — Enterprise assistant

Enterprise assistant

System
A company-wide AI assistant over internal systems and documents, with chat history and access controls.
What it does
Answers questions, finds information, and drafts content — grounded in the company's own knowledge, always with sources.
Outcome
One private place to ask, in daily use across teams — instead of hunting through tools and inboxes.

From our client work — rebuilt with no client data, numbers, or branding.

A de-identified security operations dashboard: coverage tiles, a monthly rounds chart, and incident breakdowns
Fig. 03 — Security operations dashboard

Security operations dashboard

System
A live operations view for a distributed security workforce — rounds, incidents, and coverage across sites.
What it does
Rolls patrol logs and incidents into planned-vs-done coverage, per-guard and per-zone breakdowns, and exception tracking.
Outcome
Operations leads see coverage gaps and incident hot-spots at a glance, per site and per shift.

From our client work — rebuilt with no client data, numbers, or branding.

Wide simulated velocity field with contour bands, duotone viridian on paper
Fig. 04 — Velocity field, plant-scale simulation
Pressure contours over an airfoil-like body rendered as layered duotone bands
Fig. 05 — Aerodynamic surrogate model

Aerodynamic surrogate model

System
Geometry-dependent external aerodynamics.
Physics
Compressible flow, trained on high-order CFD.
Outcome
Per-geometry predictions in milliseconds, opening design spaces that solver time previously closed off.

Capability study — illustrative, not a client engagement.


Principles

What we hold to

  1. Prove value before scaling it.

    One narrow system in production beats a platform on slides.

  2. A model must obey the domain it represents.

    When a system can break the rules of the thing it models, it eventually will.

  3. The measure is the decision it serves.

    Accuracy that changes no decision is decoration.

  4. Boring interfaces, serious machinery.

    The craft goes into what the system does, not how loudly it says it.

  5. Handover is part of the build.

    A system your team cannot own is a subscription, not an asset.


Ways of working

How we work

The principles above are the standard. This is the shape an engagement takes.

  1. A squad is the unit of delivery.

    Senior and cross-functional: physicist, ML engineer, software engineer, and an engagement lead who stays.

  2. Scrum, adapted for client work.

    Sprint cadence with your product owner inside the loop — and a working demo every sprint.

  3. Every phase ends at a stop point.

    You see working software before committing further. Continue, redirect, or stop — the call is yours.

  4. We build inside your environment.

    On your data, alongside your engineers and operators — not from a deck at a distance.

  5. Tech debt is tracked, not hidden.

    A visible register, time budgeted each sprint, and a handover with no surprises in it.


03 / About

About

Ferre Torres BV is an AI consultancy registered in the Netherlands, helping companies integrate AI into how they work — from strategy and assistants to AI-first applications.

Founded by a physicist-turned-AI-engineer, with production AI platforms and agentic systems in daily use and earlier roles spanning applied ML and AI hardware. The numerical-methods background — neural operators, high-order PDEs, and models that keep the physics inside — is what lets the team build real-time surrogates and self-learning digital twins when a problem calls for them.

We publish in the field we practice — neural operators, real-time surrogates, and the validation of digital twins — in peer-reviewed journals. Research collaborations are public; client engagements are confidential by default, named only with written consent. A Netherlands BV, serving companies across the EU, the Gulf, the United States, and Asia. The team works from the Netherlands, Spain, the Middle East, Tennessee in the United States, and Thailand, across European, Gulf, US, and Asian time zones.

The Held-Out Method

Four steps, one rule: a system is not done until it survives data it has never seen.

  1. Scope Name the system, the question it must answer, and the decision it serves.
  2. Model Build the system on your data — and on the governing physics when the system is a twin.
  3. Validate Test against data and behavior the system has never seen — and, for a twin, against held-out physics too.
  4. Deploy Ship it into your operation, instrumented, with handover your team can own.
Dotted globe in duotone viridian on warm paper, marking the team's locations in the Netherlands, Spain, the Middle East, the United States, and Thailand
Fig. 06 — Team locations, orthographic projection

Where to start

Start from your question.

The same AI core, framed for what each decision-maker needs answered.

CEO Operating impact

Turn AI pressure into a visible operating plan.

  • Which workflow should move first?
  • What result proves the investment?
  • How does this become company infrastructure?
CFO Financial intelligence

Replace reporting drag with live finance signals.

  • Where do BI reports slow decisions?
  • Which metrics need a narrative, not just a number?
  • What can be automated safely?
CTO Production architecture

Ship AI systems that fit your stack and controls.

  • Which data sources are trustworthy enough?
  • How are permissions and audit trails enforced?
  • What is the path from MVP to production?

Most engagements start with one workflow — a focused MVP or proof-of-concept — then scale what proves useful into production. Book a call

Larger programs keep their system integrator — we join as the physics and AI specialist workstream, prime or subcontract. We build on the simulation and data platforms you already run, and resell none of them: a recommendation earns us nothing but the outcome.


04 / Contact

Start a conversation.

Tell us about your system. We'll tell you honestly whether AI can help — and how we'd begin.

Book an introductory call Opens your email app, or Gmail if you don't have one set up. The address is copied either way.

Conversations are confidential, and we are glad to sign an NDA before any technical discussion. Our client list is confidential by default — as yours would be — and references are available under NDA.

We reply within one business day.