Anticipatory intelligence applied to industrial design

Evidence while your product is still a sketch.

At the conceptual design stage, when change is still cheap.

We bring the evidence forward so you can keep more options open and choose among them with data. Fewer late redesigns and fewer delays.

Keep your options open for longer.

Fig. 1

Each line is a design option; each dashed line, new evidence. Options close when the evidence arrives, not before. In the end a few remain and a person chooses one (red circle).

Illustrative
The challenge

The cost of a late decision is the whole product.

Many design decisions are made with missing or partial information, scattered evidence and consequences that remain invisible. When those consequences appear, change is already expensive, slow and uncertain.

Decisions made late

Constraints appear when modifying the project has already become expensive.

Fragmented workflows

Tools are not always connected. Teams lose context, continuity and shared judgement.

Hidden impact

Performance, environmental impact and regulatory implications remain invisible until it is too late.

In the words of design leaders

“If the aero doesn’t work, the car doesn’t work.”

Exterior Chief Designer · car manufacturer

“50–70% of projects are cancelled because the direction no longer meets business needs.”

Design Director · electric vehicle manufacturer

“That’s the biggest problem… sorting out the good stuff from the bad.”

Former Chief Creative Officer · car manufacturer

One decision platform. Evidence engines.

  • Evidence engine · Physics & aerodynamics · TRL 5 · Open to pilots

    AeroFlow

    AI aerodynamic assessment from 2D views or sketches, before CAD. It compares concept proposals in seconds and shows which ones deserve a 3D model and CFD. It does not replace high-fidelity simulation: it directs it.

    Today it works on the type of vehicle it has been trained on. We adapt it to your range with your simulations.

    See AeroFlow →Evidence and limits →

  • Evidence engine · Geometry · TRL 7

    Neural 3D reconstruction

    Reconstructs the geometry of a product from images, views and scans, enabling reasoning about form, structure and physical behaviour without relying on complete CAD models.

    See neural 3D reconstruction →

  • Decision platform · In development

    Design Engines

    Brings together alternatives, conflicting criteria and constraints, and establishes what evidence each choice needs.

    See Design Engines →

  • The next evidence engines are open to industrial collaboration for joint research and development.

Before building the 3D

Aerodynamic feedback arrives while the form is still open.

Classic industrial development flow
  1. Concept views
  2. Develop the 3D model
  3. Set up and run CFD
  4. Feedback reaches the design → redesign
With AeroFlow
  1. Concept views
  2. AeroFlow prediction
  3. Concept review
  4. 3D development and CFD only for the selected concepts
A typical case

A hundred sketches ten selected five 3D models.

Solid frame: with aerodynamic evidenceDashed frame: without evidence

100

all of them, with AeroFlow, in a fraction of the time

Every one of the hundred sketches gets an aerodynamic estimate before anyone decides which ones move forward.

Today

5–8.5

days per concept

Aerodynamic evidence arrives after the 3D model. The other 95 concepts are discarded without it.

With AeroFlow

1–5

minutes per sketch

All 100 sketches, within a day of computing. 3D and simulation are kept for those that deserve it.

Possible saving

40%

less 3D modelling

If aerodynamic evidence rules out two of the five models.

Time from sketch to a first aerodynamic answer

Working time per concept, logarithmic scale. Today: 3D modelling plus simulation set-up. AeroFlow: view preparation plus prediction.

Assumptions behind this case

Process figures taken from automotive design studio practice; they vary by programme. AeroFlow: automatic preparation of three views from the sketch (20 to 90 seconds per view) and prediction in under 2 seconds per view. The estimate is for screening concepts; it does not replace high-fidelity simulation.

Estimate based on real practice figures

Built for those who have to decide.

Strategic clarity

For leadership

Gain early visibility into cost, risk, performance, time and impact before decisions become commitments that are difficult to reverse.

Feedback in time

For design

Get technical feedback while the form is still being explored, without waiting for the 3D model or the simulation result.

Technical confidence

For engineering

Explore alternatives faster, anticipate performance and bring aesthetic intent closer to technical feasibility.

How we think →

Tested with design teams. Backed by research.

  • Trials

    Trial users

    Automotive design teams in industry, academia and design services have tried AeroFlow. Names are kept confidential.

  • Research

    Peer-reviewed publications

    Work presented at SIGGRAPH Asia 2025 and PhD Award at EACWE 2025.

    See research →

  • Ecosystem

    Institutional backing

    Spin-off of the Universidade da Coruña · CITIC · NVIDIA Inception.

    See company →

We bring intelligence to the start of the process.

For product teams, industrial partners and institutions.

weare@deepdesignsystems.com