A Chat With Gary Brotman, CEO of Secondmind on Building Data-Efficient AI Software

Tell us about Secondmind

Secondmind is a UK deep-tech scaleup building data-efficient AI software for model-based engineering that reduces complexity in physical product design. Today we operate in Europe, Japan and North America, with a primary focus on the automotive sector, where modern vehicles combine tightly coupled mechanical, electrical and software systems with billions of potential design and calibration combinations. The scale of this rising complexity now exceeds what traditional simulation and testing workflows can manage efficiently.

Secondmind acts as a cognitive layer between engineering intent and the physical systems being developed. It integrates into existing toolchains and identifies the minimum set of simulations and physical tests required to meet a defined performance objective. Rather than replacing engineering expertise, we amplify it by helping engineers discover new design options, make better decisions earlier, and reduce costly late-stage rework.

This translates into measurable bottom-line impact. For example, a major car manufacturer recently used Secondmind’s Design Space Exploration capabilities to discover new e-motor designs in less time than their incumbent tools, while simultaneously reducing bill-of-materials cost. By intelligently navigating trade-offs between performance variables, Secondmind helped engineers select a design that exceeded power and driving comfort targets while reducing component cost by an estimated $27 per unit — equivalent to approximately $11 million in annual savings at full production scale. This combination of time savings, performance improvement and cost reduction defines our value proposition.

What makes Secondmind unique?

Much of today’s AI narrative focuses on scale — larger models, more data and more compute. Engineering environments operate under fundamentally different constraints. Simulations consume significant computational resources. Prototypes are expensive. Physical testing takes months. In this context, efficiency is not optional; it is structural.

Secondmind was built around data-efficiency. Our Active Learning system reduces required simulation and test data by up to 80% while maintaining high-precision models with calibrated uncertainty. Rather than running thousands of experiments in a brute-force manner, we identify and execute only those that are most informative.

In calibration, this means our software can optimise control system performance in roughly half the time of incumbent tools. That reduction in iteration cycles materially compresses development timelines.

In addition to data-efficiency, a key differentiator is our ability to blend virtual simulation data with real-world test data within a single probabilistic framework. Most AI systems in engineering rely either on historical datasets or purely simulated environments. In complex physical systems, neither is sufficient alone. Simulation provides breadth; physical testing provides ground truth. By fusing both, Secondmind enables efficiency without sacrificing real-world fidelity. We view this hybridisation as essential to scalable and trustworthy AI in precision engineering.

How has Secondmind evolved?

Automotive was a deliberate beachhead choice to validate our broader platform strategy. Our belief was that if we could deliver measurable outcomes in one of the world’s most complex engineering environments, the same technology approach would translate to other industries with expensive experimentation and constrained design spaces. Our success in automotive has validated that thesis.

We focused initially on vehicle powertrains and control systems, where the engineering trade-offs are acute and the business value of faster, more confident decisions is quantifiable. Our work with global OEMs and Tier 1 suppliers demonstrated repeatable economic value inside real R&D and production workflows. This strengthened our conviction that building a scalable platform from a focused vertical entry point was the right technology and business strategy.

What can we see from Secondmind in the future?

Our future growth follows both vertical and horizontal paths.

Vertically, we will extend further across the automotive lifecycle — from early-stage system design and calibration into manufacturing optimisation and continuous in-field performance improvement. The same data-efficient framework that accelerates design exploration can also improve production yield and operational performance once vehicles are deployed.

Horizontally, we will apply the platform to industries facing similar engineering complexity challenges, including semiconductors, advanced materials, consumer appliances and data centre infrastructure. We view the modern data centre as a stationary powertrain, which is a tightly coupled system responsible for energy conversion, distribution, thermal management and performance optimization at scale. The rapid build-out of AI infrastructure is placing increasing strain on power generation, distribution and cooling systems, each a complex physical engineering challenge in its own right. Improving their efficiency demands the same kind of data-efficient design and performance optimisation that Secondmind cloud-native software has already demonstrated in automotive powertrains and control systems.

The common denominator, however, is not the sector itself, but the presence of constrained engineering environments where blending simulation with real-world data creates structural and competitive advantage.
Longer term, we see Secondmind as a foundational player in Lifecycle AI – AI that is applied and traceable across the entire product lifecycle, from early discovery through design, production and real-world operation of systems in the physical world.

Our overriding purpose remains constant: not to replace engineers, but to augment their intelligence in environments where precision, efficiency and trust are non-negotiable.