Intelligent Design
Jarek Rzepecki on how AI can help design complex hardware as a single, unified system
When the electric motor was first invented, it didn’t immediately improve productivity in factories that had previously relied on steam engines. As steam engines were so large and their power distributed mechanically, their setup determined how electric motors were integrated into workflows. Electric motors were initially used to replace a central large source of power — but it was later discovered that the efficiency gains and benefits really started when that central source was replaced by a larger number of smaller motors.
Such is often the case in innovation. We might have the means of improving processes but it takes time for them to be properly adopted and adapted to their intended uses. This is particularly true in Britain’s £747 billion engineering industry, where breaking out of the conventional way of operating can be difficult.
Engineers currently design individual components as best they can, and then piece them together into a complex system later. But that approach is facing challenges driven in part by resource scarcity, but also a desire to achieve greater efficiency. The solution may be to reimagine our whole approach to design by thinking of complex hardware as a single, holistic entity that can be designed and optimised at the same time.
Jarek Rzepecki, the new CEO of Monumo, believes this is the foundation for all engineering in the coming decades. By moving away from component-led design toward a system-level optimisation, Monumo is showing that significant cost reductions can be made without sacrificing performance.
If Monumo succeeds, it will enable a leap in industrial productivity without the energy and infrastructure requirements of traditional AI models. I sat down with Jarek to discuss his transition from CTO to CEO, the systemic challenges facing the UK’s deep-tech investment landscape, and how Monumo is leveraging foundational physics and its Anser AI computation engine to radically optimise the design and sustainability of electric motors.
What we discussed
Jarek’s transition from CTO to CEO and how to balance technical rigor with investor and customer relations.
The shift from component-based engineering to holistic, AI-driven system design.
Why the UK’s focus should move from data centres for large language models towards the intersection of AI, physics and chemistry.
The structural opportunity for AI-native manufacturing in the UK.
Why the path to UK productivity lies in building new industries from scratch rather than retrofitting incumbents.
Lessons for scaling deep-tech hardware companies
For policymakers:
Prioritise ‘AI for Science’ over LLMs. The UK doesn’t have a comparative advantage for LLMs considering the energy and infrastructure required. Instead, the UK should focus on the intersection of AI, physics and chemistry, where the UK has a distinct research advantage and lower infrastructure costs.
Build AI-native manufacturing. Rather than trying to turn around incumbent industries, the UK should create new, AI-native industries that are built on the latest techniques from day one.
Lighten the grant admin burden. Set a high bar to win grants, but reduce the ongoing reporting and administrative load once awarded — including for any new schemes. The current burden falls heaviest on the small companies grants are meant to support.
Fix the London listing environment. The current cost-benefit analysis makes it unwise to list in London. Improving the appeal of the London Stock Exchange would provide a benefit to the economy that far outweighs any loss in immediate oversight or the administrative costs of transition.
Fund fundamental AI research. UK grant structures currently require AI research to be tied to manufacturing applications, making it nearly impossible to fund pure AI research.
For founders:
Prepare for disruption. Deeply disruptive technologies face a credibility gap. Founders must be prepared to build trust and credibility over time as incumbents get comfortable with new, AI-native ways of doing business.
Focus on resilience. Building a company is never easy. Beyond technical curiosity, the tenacity and creativity to build what hasn’t existed before are the two primary qualities of a successful entrepreneur.
Own the full stack. While this is not possible for many businesses, for founders in the deep-tech or AI space, building your own foundational models rooted in first principles rather than building on top of someone else’s API can offer tremendous advantages.
Prove yourself in one market, then diversify. Cross the Valley of Death with radical focus before expanding. Once your core technology is proven, diversification across sectors becomes a strength — especially when the underlying technology is horizontally applicable.
Full interview
I. Building Monumo
Monumo applies foundational physics to hardware architecture. Can you explain what you’re building and why it matters?
Monumo is an engineering AI company focused on creating complex engineering designs, at speed and ready for production. We are currently working with global brands in the automotive, domestic appliance and industrial sectors, initially focused on electric motor systems, with robotics and aviation next on the roadmap.
Monumo’s Anser® Engine uses deep-tech and AI to explore millions of potential design options, evaluates them against set customer parameters and objectives and selects the optimal designs for production in just hours.
The Anser Engine does not require customer data to create potential designs but just the requirements it must fulfil, such as shape, size, torque, efficiency or cost. After understanding the requirements, the Anser Engine generates its own data based on high-precision simulations and then assesses each design for fit. It generates results at incredible speeds reaching beyond human capabilities. So what we’re building is the foundation for how all engineering will be done in the coming decades, such as for cars, planes, industrial equipment and energy turbines.
Can you give me some examples of projects you’ve worked on?
In one recent project, the Anser Engine revealed a 17% cost reduction over a customer’s original electric motor design, enabling savings of up to €70 per motor.
In terms of how that 17% cost reduction was achieved, in an electric motor, there are three main materials: electrical steel for the stator and rotor, copper for the windings, and permanent magnets. Of all these materials, the magnets are by far the most expensive. In this particular project, we managed to reduce the required volume of permanent magnets. By moving the magnets around, changing the shapes of air pockets, and adjusting the design, we achieved a motor with the same performance as the original but with a significantly lower material cost. It also helps reduce dependency on the volatile rare-earth supply chain.
When you consider that the top 10 automotive OEMs [Original Equipment Manufacturers] will need many new motor designs in the next decade, the potential for industry-wide savings is staggering.
Aside from electric motors, what else are you looking to optimise?
At the moment, we are working exclusively with electric motors, but across very different sectors. A motor for a washing machine is quite different from one for an EV or a drone. They use the same underlying physics, which is why our technology works for all of them, but the metrics for success differ.
In drones, weight is the priority. Every gram saved is a gram the drone doesn’t have to lift. For industrial applications, efficiency is key. These motors often run 24/7, so even small electricity savings add up to huge cost reductions. In the EV space, cost is the main driver to help manufacturers compete. In humanoid robotics, there is a strong need for dedicated actuators. Currently, companies either buy expensive high-quality actuators or hobbyist-grade ones that aren’t good enough. Monumo can produce designs for dedicated actuators that fit the specific requirements of humanoid robots.
Our vision is to be the go-to designer across all these verticals. Diversification is important because sectors have golden eras. Eight years ago, EVs were moving extremely fast; now, that has slowed down while robotics is kicking off. If you put all your eggs in one basket, you risk ending up in trouble. Since our technology applies across the board, doubling down on just one sector would mean missing out on revenue streams in white goods or industrial motors. Addressing as many markets as possible is the right choice for risk reduction.
In the distant future, we would like to be optimising the entirety of engineering. But optimising a whole car is orders of magnitude more complex than optimising the powertrain.
Our expansion will be gradual. If you take the drone example, we currently optimise the motor, controller, and inverter together. The next step could be optimising the shape of the propeller. We might not optimise the frame yet, but we’ll take it step-by-step. If we see a benefit at one stage, we move to the next. While a full drone is within our reach, we probably won’t be optimising car seatbelts or entertainment systems anytime soon!
What has enabled this transition from designing components separately to engineering the system as a whole in the last few years?
The big shift is around the speed of compute and the scalability of our Anser Engine. It’s also about having the right focus. In one recent project, we evaluated 3,000,000 designs in just three days. That level of iteration is simply impossible for human engineers using traditional methods.
If you keep doing that over the next few decades, we’ll end up being able to design an entire aircraft or car as a single entity. That’s really exciting because it will transform the performance and use of materials in so many products that are essential to modern society.
To power that kind of transformation, I’d imagine you can’t just rely on standard off-the-shelf LLMs. Do you train your own AI models, or use off-the-shelf ones?
For what we do, there are no off-the-shelf models, though we do use known architectures. Models have to be trained by us, from scratch, and we own the full stack. One thing to mention about AI for Science or Engineering is that it often does not require as much compute or the energy that frontier LLMs do. I really don’t think it’s the best use of resources to chase after the frontier LLMs in the UK, because we don’t really have a comparative advantage when you consider the energy and infrastructure required. We can get ahead in AI for Science or Engineering without those same requirements, especially if you leverage deep scientific and engineering expertise the UK does have.
You recently transitioned from CTO to CEO. Now that you’re stepping into the lead role, what different skill sets do you think this will take?
Monumo is a deep-tech company, so a huge portion of running the company was always centred around the technology. The development team is the core of the company. To some degree, I’ve been doing a lot of this since the beginning. I have also been heavily involved in VC conversations and customer interactions in the past.
In terms of skill set, I don’t feel there is a huge gap between what I have been doing and what I will need to do going forward. However, the ratios will change. In the past, I might spend 30% of my time with customers, 60% with technology, and 10% with VCs. Now, I imagine it will be a more even split of a third on each of these. I still want to stay close to the technology — it’s vital that I remain close to our technical developments — but being close to customers and VCs is equally important as we look to raise more funds. So I’d say the skills remain similar, only the focus shifts.
Tell me a bit more about Monumo’s history. How did it begin, what’s your team like, and has anyone else attempted to build something similar?
The idea came from research programmes at Arm. The founding team noticed that electric motors were underdeveloped and saw scope for improvement.
In terms of the team itself, we have quite a few Arm veterans in the business. Over half of the team have PhDs in physics, electronics, machine learning, and computer science. It’s a very science-based team in Cambridge, and then we have a team in Coventry that has the domain expertise in automotive engineering and prototyping.
Regarding competitors, there are three main groups, but no one does everything we do. There are incumbent simulation companies such as Ansys and Siemens. They have high-fidelity simulators, but their approach is very human-centric. Their stacks weren’t built from the ground up for AI. Then there are AI-for-Physics startups, which are trying to crack AI for engineering but they often lack access to high-quality simulation or in-house engineering expertise. There are also engineering giants who have huge domain expertise but lack the specialised simulation or AI expertise.
Monumo sits in the middle of those three. We combine domain expertise via our hardware team in Coventry, our own high-quality simulator, and an AI team that uses that data to train models. To my knowledge, we are the only ones with all three pieces.
II. Navigating the UK’s industrial investment landscape
What has been your experience with the UK’s funding landscape?
R&D tax credits have been enormously helpful, and I hope they don’t change. But the grant process is quite onerous. I think there should be a high bar for getting grants in the first place, but once the grant has been made, the administration and reporting should be less onerous. At the moment, they put a huge, continuing burden on businesses.
One of the issues is that emerging technologies such as AI are so deeply disruptive to long-established companies that there is a natural timidness and lag around adopting them. Support to help large companies start the transition and the evolution towards an AI-native way of business would be helpful.
You have spoken about the UK’s exit problem. What needs to change?
At the moment, I don’t think I could find a single adviser who would say that if we were to float the company, we should do it in London. It comes down to a fundamental valuation gap. If you look at multiples paid on the Nasdaq versus the London Stock Exchange (LSE), a deep-tech company often sees a 30–50% discount just for being listed in the UK. That’s largely because we lack a dense ecosystem of institutional investors and analysts who truly understand deep-tech timelines. Most British funds are looking for immediate dividends or steady cash flow, whereas a company like Monumo requires a decade-long view.
Liquidity is also a massive issue. There aren’t enough anchor tech investors in London to keep the stock price stable during the early years of a listing. For a founder, if you list in London you just aren’t getting the visibility or the capital you need to compete with a Silicon Valley peer.
When a British startup reaches a certain scale, they often face a choice: selling to an American or Chinese giant, or doing an IPO in London. From a government perspective, a trade sale provides a quick, one-time shift in revenue via capital gains tax and a clean exit for early investors. But the country loses the IP, the high-paying head-office jobs, and the future tax base of a trillion-dollar company. I’m arguing that we should prioritise the long-term health of the LSE over those short-term wins.
III. Conclusion and advice
If you were advising the Secretary of State for Science, Innovation and Technology, what would be your priorities?
It would be good if the government talked a little less and did a little more. There is a lot of talk about AI sovereignty, but that needs to translate into real help for tech companies. It’s still significantly easier to access funds in the US than in Europe.
One specific issue is government grants. It is very hard to find grants that support pure AI research; they usually require you to link it to manufacturing. Fundamental research is the bottom of everything. Companies like OpenAI or Anthropic are software companies with heavy foundational AI components; the UK doesn’t really have an equivalent domestic AI powerhouse. This is a strategic problem. AI is a strategic resource, and relying on other countries for it is dangerous.
What is your advice to founders in this space?
I wish this wasn’t the case, but you have to be extremely resilient to cope with building a company. It’s never easy. You also need more creativity than is often thought — you really need to be able to build something that hasn’t existed before.
You have to be prepared to cross the Valley of Death into more fertile territory. You also need to maintain a radical focus, prove yourself in one market first before you get distracted by the multiple potential applications of a deep-tech solution.
We ask all our guests the same closing question: what’s one interesting thing you’ve read or listened to recently that you’d like to share?
Recently, I read The Magic Machine: A Handbook of Computer Sorcery by A. K. Dewdney. It’s a delightful playground for technical minds who want to rediscover the joy of computing. It covers a surprisingly broad range of topics from chaos theory to early graphics, all paired with hands-on programming challenges. The author has a rare gift for making deep computational ideas feel like puzzles you actually want to solve. For me, it was a trip down memory lane, revisiting how computer programming was in the days when I was growing up and learning it for the first time.