Key Learnings from Episode 19: Scaling AI in Heavy Industry with Dr Noah Miller, COO of Gigaton

In Episode 19 of Beyond the Raise, Alex speaks with Dr Noah Miller, COO of Gigaton, about what it takes to deploy AI inside some of the world’s most complex industrial environments. From decarbonising cement production to scaling a venture-backed team after Series A, Noah explores why focus, technical credibility, trust and hiring discipline are essential when building technology designed to control the physical world.

Using AI to Change How Heavy Industry Operates

Gigaton exists to change how heavy industry is run, beginning with cement.

Cement production accounts for around 8% of global emissions, yet it remains fundamental to the infrastructure we rely on. The challenge is therefore not simply to replace cement, but to find ways of producing it more efficiently and with fewer emissions.

Gigaton is building an AI control system designed for energy-intensive industrial processes. It combines machine learning with industrial expertise, software engineering and control theory to improve how plants operate.

For Noah, this brings together different parts of his own career. He began in academia with a PhD in chemical engineering, moved into renewable energy modelling with Advanced Infrastructure, and later joined what was then Carbon Re as Head of Machine Learning before becoming COO.

Key Takeaways

• Heavy industries such as cement, steel and glass remain essential but highly energy intensive.

• Gigaton is using AI to modernise the control systems behind these industrial processes.

• Deep industry knowledge needs to sit alongside machine learning and software expertise.

• Noah’s role has evolved from building machine learning capability to helping scale the wider organisation.

Why Cement Offers an Immediate Decarbonisation Opportunity

Gigaton chose cement partly because of the scale of its emissions, but also because there is an opportunity to make an impact using plants that already exist.

A significant proportion of cement emissions comes from the chemical reaction involved in turning limestone into the material required for cement. Another part comes from the enormous amount of energy required to heat a cement kiln.

Gigaton focuses on the latter. By controlling the production process more efficiently, plants can burn less fuel while maintaining the required product quality.

That creates an important alignment between climate impact and commercial value. Reducing fuel consumption cuts emissions, but it also reduces one of a cement producer’s significant operating costs.

Key Takeaways

• Cement is responsible for around 8% of global emissions.

• Existing cement plants will continue operating, making improvements to current processes important.

• More efficient control can reduce the fuel required during production.

• In this case, the environmental and commercial incentives are closely aligned.

Small Efficiency Gains Can Have a Huge Physical Impact

When industrial processes operate at enormous scale, relatively small percentage improvements can translate into substantial savings.

Noah describes one European customer where Gigaton achieved approximately a 1.5% reduction in fuel consumption during an evaluation. According to Noah, that equated to roughly €1.2 million in annual savings and around 11,000 to 12,000 tonnes of CO2.

This scale is part of what excites him about applying machine learning to heavy industry. Software written by engineers in one location can directly change the behaviour of enormous pieces of industrial equipment somewhere else in the world.

Gigaton describes this distinction as moving atoms rather than bits. The outcome of the software is not confined to a screen. It changes a physical process.

Key Takeaways

• Small efficiency improvements become significant when applied to large industrial processes.

• Commercial savings and carbon reductions can happen simultaneously.

• Industrial AI creates a direct connection between software and physical-world outcomes.

• Measuring impact in fuel, cost and carbon keeps the technology connected to tangible results.

Moving Further Down the Control Stack

Gigaton did not arrive at its current product immediately.

Its first approach was an open-loop system that recommended changes to plant operators. The operator would see a suggested value and manually enter it into the existing control system.

The problem was that recommendations were not always followed. Operators could disagree with them, miss them or simply have other priorities at the time.

Gigaton then moved further down the control stack, setting targets for an existing automated control layer. That exposed another problem: the underlying control technology was not always capable of implementing Gigaton’s recommendations effectively.

Eventually, the team concluded that it needed to control more of the process itself. Rather than simply advising an existing system, Gigaton began building the control and optimisation layers together.

Key Takeaways

• Technically correct recommendations are only valuable if they are implemented.

• Human intervention created limitations in Gigaton’s first approach.

• Working through existing control systems introduced another layer of constraint.

• The product evolved towards deeper control of the industrial process.

Focus Sometimes Requires Giving Something Up

Moving towards the new control system required Gigaton to make a difficult commercial decision.

Noah explains that the company reduced its revenue by around 50% at the beginning of 2026 because some customers still wanted the previous product.

Continuing to serve those customers would have generated revenue, but it would also have divided the company’s attention. The leadership team believed the deeper control system was the product capable of creating the impact and scale they were aiming for.

The decision gave the team greater clarity. Instead of supporting multiple versions of the future, the company could concentrate its resources on one direction.

For Noah, intensity at a start-up should not simply mean working more hours. It can also mean intensity of focus, where the organisation is clear about the few priorities that genuinely need to happen.

Key Takeaways

• Existing revenue can become a distraction if it belongs to a product the company no longer believes in.

• Gigaton accepted a significant short-term commercial impact to create greater strategic focus.

• Clear priorities help teams understand where their effort should go.

• Intensity can come from focus rather than simply longer working hours.

Customers Helped Give Gigaton Confidence to Go Deeper

The decision to replace more of the existing control layer was supported by both internal and external confidence.

Some of Gigaton’s key customers shared its belief that industrial control needed to change more fundamentally. These development partners were prepared to work closely with the company as it moved deeper into the plant’s control systems.

Building those relationships required more than AI expertise. Gigaton needed people who understood cement deeply enough to have credible conversations with plant operators and industry leaders.

Internally, confidence came from the team itself. Noah describes a group of highly capable people who believed the problem could be solved even though it had not been done before and the timescales were demanding.

Key Takeaways

• Strategic customers can become development partners rather than simply buyers.

• Customer confidence helped validate Gigaton’s move towards deeper industrial control.

• Industry expertise is essential when introducing new technology into specialised environments.

• Internal belief and external validation reinforced each other.

Series A Increased the Pressure Rather Than Reducing It

Gigaton's $26 million Series A provided substantial resources, but Noah describes the funding as the beginning of another challenge rather than a moment to relax.

The company established a simple sequence. In 2026, the new product needed to work. In 2027, it needed to scale.

That meant building more quickly while simultaneously growing the organisation capable of delivering it.

Gigaton began 2026 with around 22 people. At the time of recording, Noah says the team had grown to around 32, with plans to reach approximately 50 by the end of the year and potentially 70 to 80 by the end of the following year.

The funding also reassured customers. Large industrial businesses considering rolling Gigaton out across dozens of plants needed confidence that the company would still be there to support them years into the future.

Key Takeaways

• Fundraising creates new expectations alongside new resources.

• Gigaton separated its challenge into making the product work and then making it scale.

• Rapid company growth creates organisational challenges alongside technical ones.

• Financial backing can give enterprise customers greater confidence in a supplier’s longevity.

Industrial AI Is Very Different from Generative AI

AI in a cement plant bears little resemblance to asking a large language model to draft an email.

Gigaton needs data from thousands of sensors operating across complex industrial environments. That data can be messy, equipment can be old and connectivity can be inconsistent.

There is also far less existing data showing how to control a cement plant than there is data available for common generative AI tasks.

Gigaton therefore combines customer data with its understanding of cement production, physical chemistry and industrial processes to build models capable of operating across different plants.

The objective is still to build a repeatable product rather than a bespoke consultancy solution for every customer.

Key Takeaways

• Industrial AI operates with very different data and constraints from generative AI.

• Industrial environments can produce incomplete, inconsistent or difficult-to-access data.

• Domain knowledge helps compensate for the lack of large public datasets.

• The commercial challenge is creating a repeatable product that can adapt to different physical environments.

Trust and Explainability Are Essential

Giving software influence over critical industrial equipment requires a high degree of trust.

Gigaton builds that trust partly through industry expertise. Its teams understand cement and can speak directly to the engineers and operators responsible for running plants.

The company also spends significant time on site. Solutions engineers work alongside operators in control rooms, including night shifts, to understand how the plant behaves and demonstrate how Gigaton's system works.

Explainability has become a fundamental part of the product. Noah says the company initially underestimated its importance, assuming that strong performance would be enough.

It was not. If operators cannot understand why the system has made a decision, they can simply turn it off.

Key Takeaways

• Technical performance alone is not enough to earn trust in critical industrial environments.

• Domain expertise helps technology teams build credibility with plant operators.

• Spending time with the people actually using the technology reveals problems that can be missed remotely.

• Explainability is a functional requirement because operators need confidence in the system’s decisions.

Scaling Means Teaching Other People to Hire Well

When Gigaton was around 20 people, Noah and CTO Bob Gregory could be directly involved in hiring almost every engineer.

That model cannot survive rapid growth.

As the company has expanded, team leads have taken greater responsibility for recruitment. Noah’s job has therefore shifted from personally making every hire to creating the standards and processes that allow other people to hire effectively.

Onboarding and probation have also become more formal. Noah remains involved with new starters, both to build relationships and to make sure expectations are understood early.

The goal is not process for its own sake. Every new system needs a clear reason behind it.

Key Takeaways

• Founder and executive-led hiring becomes unsustainable as headcount grows.

• Hiring standards need to become transferable across different managers.

• More structured onboarding helps communicate expectations as the organisation scales.

• New processes are easier to accept when people understand why they exist.

Scaling Requires Process Without Losing the Start-up Mentality

Growing from start-up to scale-up inevitably introduces more structure.

For Gigaton, the challenge is introducing what the company needs without creating unnecessary bureaucracy.

Some processes are unavoidable. Hiring needs consistent assessment. Expenses need controls. Managers need clearer ways of onboarding and developing people.

Noah sees part of his responsibility as explaining the value behind these changes. Structured interview scorecards, for example, are not administrative exercises. They help candidates receive a more consistent and fair assessment.

At the same time, Gigaton wants to retain the speed, ownership and willingness to solve difficult problems that characterised the smaller company.

Key Takeaways

• Scale requires more structure, even when a company values autonomy.

• Processes work better when teams understand the problem they are designed to solve.

• Hiring systems become increasingly important as more managers become involved in recruitment.

• The challenge is adding enough structure to scale without creating unnecessary bureaucracy.

High Standards Do Not Have to Mean Long Hours

Gigaton describes its culture as an adult culture, built around high standards alongside individual flexibility.

Noah is clear that measuring commitment by how many hours somebody spends sitting at a desk is not the same as maintaining high standards.

Instead, the company tries to focus on effort, care and outcomes.

Different people achieve that in different ways. Noah discusses working around life with his young family, while CTO Bob Gregory might spend part of a weekend absorbed in a difficult coding problem and then take time during a weekday to walk his dog.

The important thing is that flexibility exists alongside trust and a shared expectation that the work will be delivered to a high standard.

Key Takeaways

• High standards should be measured through the quality of work rather than performative hours.

• Flexibility can coexist with demanding expectations.

• Leaders can model different ways of working rather than prescribing one pattern.

• Trust becomes increasingly important as a company grows.

Hiring After Fundraising Needs More Than One Route

Gigaton’s rapid growth has also changed how Noah thinks about talent acquisition.

The company now has an internal talent function, but it continues to work selectively with specialist recruiters.

Noah approaches the decision pragmatically. Some roles are sufficiently specialised that an external recruiter with deep knowledge and an established network can add significant value. There are also limits to how much an internal team can handle when many roles are open simultaneously.

Personal relationships matter here. Noah values having a trusted network of people he can approach when the business needs specific external support.

Key Takeaways

• Internal talent teams and specialist external recruiters can play complementary roles.

• Specialist searches may benefit from recruiters with established market knowledge and networks.

• Internal hiring capacity becomes a practical constraint during rapid growth.

• Trusted relationships make it easier to access external support when it is genuinely needed.

Hiring Quality Cannot Be Sacrificed for Speed

Some hiring lessons only become clear after getting them wrong.

Noah points to the idea of maintaining a high talent bar. It sounds straightforward until a company compromises on a candidate who appears good enough and then spends months managing the consequences when the hire does not work out.

Reference calls have therefore become particularly important to him. Interviews provide only a few hours of exposure to somebody, whereas former colleagues may have worked alongside that person for years.

Gigaton also places significant emphasis on its operating principles interview, which Noah still participates in.

When asked to rank cultural fit, quality and speed, his order is clear: cultural fit first, quality second and speed third. The talent function’s job is to deliver the first two as quickly as possible, not sacrifice them for the third.

Key Takeaways

• Lowering the hiring bar can create far greater costs after somebody joins.

• References provide a different perspective from interviews alone.

• Gigaton assesses operating principles explicitly as part of its hiring process.

• Speed matters, but not at the expense of cultural fit or candidate quality.

Looking Ahead

Gigaton’s ambition extends well beyond its current cement deployments.

At the time of the conversation, Noah says the company is working with around ten customers globally and is targeting around 50 by the end of 2027. It is also beginning to explore a second heavy industry where its technology could solve similar control problems.

The broader ambition is reflected in the company’s move from Carbon Re to Gigaton. Its mission is not limited to improving a handful of cement plants, but to achieve carbon reductions measured in gigatonnes across multiple energy-intensive industries.

For Noah, that means building a company capable of operating across hundreds of plants while retaining the technical depth and sense of purpose that got it this far.

Final Thought

Noah’s experience since Gigaton’s Series A has reinforced one lesson in particular: scaling needs to be prepared for earlier than it feels necessary.

His own reflection is that he spent much of the first half of 2026 concentrating on customer delivery when he could have invested more time earlier in building the talent capability required for the next phase.

By the time a growing company urgently needs those people, it is already late to start building the system that will find them.

For founders approaching the same stage, the lesson is to treat organisational scale with the same seriousness as product scale. The technology may create the opportunity, but the team has to be ready to deliver it.