By: Svetlana Khachiyan 

He chose an unpopular but very challenging profession and is now changing the future of industry. Sasha Malakhov founded a company that develops machine learning models for the metallurgy sector and has already achieved impressive results for its clients — giants of the industry.

Today, he shared with us his insights on the growing interest of businesses in AI, the factory without employees, and the upcoming revolution in AI and business analytics. He also talked about how advanced IT solutions help achieve impressive financial results and strategic goals, taking companies to a whole new level of development.

1. Sasha, what inspired you to create a company that provides machine learning models for the steel industry? Why did you choose this unpopular field?

My love for heavy industry is in my blood. I come from a family of engineers. On my father’s side, my relatives are involved in energy and oil extraction, while on my mother’s side, they are metallurgists. As a result, I received both engineering and mathematical education, along with experience in business transformation in heavy industry at McKinsey & Company. Over time, I noticed a niche — artificial intelligence in metallurgy — and saw strong interest from potential clients. At the same time, there was limited competition in the market, and my experience at the intersection of these two fields turned out to be valuable. I didn’t hesitate for long and decided to seize this opportunity.

2. You claim that your machine learning models have no analogs on the market. What exactly makes your solutions unique for the steel industry?

Developing such models is only possible with a deep understanding of metallurgy and expertise in modern mathematical modeling methods. The uniqueness of our solutions lies in the personalized approach we take when developing algorithms. Instead of adapting existing algorithms to the input data, we create them from scratch to ensure they reflect the physical and chemical processes occurring during steelmaking as accurately as possible.

In other words, our solutions are the result of combining deep expertise in metallurgy with advanced skills in artificial intelligence, machine learning, and optimization technologies.

3. How do you collaborate with clients? How is your technology implemented, and how do you provide post-implementation support?

We treat all our clients as partners, collaborating to address business challenges, set priorities, and explore innovations in AI and metallurgy. Our model blends strategic consulting, B2B SaaS, and software studio, drawing from the premier practices of each.

We begin by working on-site, engaging with engineers, and diagnosing business processes before developing or implementing a solution. Together, we define objectives, select the right product from our line, and customize it considering the specifics of the client’s technologies and business processes. Then, we implement the solution into current processes and train personnel for a smooth transition to the new work model. After the project is completed, we continue to support clients, adding new functionality as needed and adapting the tool to changes in their processes.

4. Your technologies are proprietary and closed to competitors. How do you ensure their security, and how can clients be sure of their effectiveness?

Calculating and tracking results is a core feature of our products. Clients can see the results on a dashboard where the data is compared with their historical performance. In terms of security, we maintain high standards at every level. Clients’ data is stored on servers located within their country, and we create isolated architectures or on-premises solutions tailored to each client. Our algorithms are monitored by humans, and all solutions undergo expert review. More advanced models operate in a closed-loop system, where experts simply oversee the process. For these solutions, we offer dashboards and emergency alerts via Teams or Slack: if critical deviations occur, clients are notified to verify the data and, if needed, switch to manual control.

5. What is involved in the process of “adoption” of your solutions?

You can think of it like a motorcycle with a sidecar, where the motorcycle represents the digital product. First, Metal Minds takes the motorcycle for a spin (testing phase), showing the client that it works. Then, we hand over the controls to the client and observe from the sidecar as they take charge. Finally, Metal Minds watches from the sidelines, ready to step in if assistance is needed.

More precisely, we integrate into the existing business process, actively manage the model, and gradually transfer control to the user. We then document the business process and the model’s use. While this adoption process may take longer, it significantly enhances the likelihood of a successful and sustainable implementation.

6. Which processes in steel production can be improved with your solutions? How does AI help improve productivity, reduce costs, and improve product quality in metallurgy?

We focus on three areas: procurement, planning, and operational initiatives, where numerous interchangeable options make it difficult for humans to account for all constraints and find optimal solutions. Our algorithms work continuously to provide optimal results.

They offer two key benefits: faster, more optimal route identification and freeing up expert time for more complex tasks that require creativity and experience.

In operational initiatives, we focus on accurately measuring chemical composition and physical properties. For example, IntelScrap determines copper content in scrap metal. When copper levels are unknown, engineers often use large safety buffers. Our model shows that copper content is typically low, with high levels occurring only occasionally. It identifies these cases and automatically adjusts the recipe, reducing safety buffers in most situations. This increases the use of post-consumer scrap and enhances circular economy performance.

7. What is your main advantage over competitors? How does the combination of consulting and software solutions help clients integrate AI into their business processes?

Metal Minds doesn’t just provide an off-the-shelf solution; we implement the solution and carry out full adoption and support of business processes to ensure the software doesn’t remain underutilized within the organization. We combine the premier of consulting and software companies — consulting for full adoption and deep integration into business processes and software companies for full technical support over the years and enterprise-level technological solutions.

8. Many companies face difficulties when implementing AI. What challenges do your clients face, and how do you help them overcome these challenges?

The three key challenges are selecting the right tools, implementing them, and ensuring adoption.

Choosing the right tools requires careful consideration. For example, GenAI isn’t suitable for chemical process modeling, and neural networks aren’t effective for linear equations. Mistakes here can undermine the implementation. Metal Minds is experienced in various modeling methods and understands their strengths and limitations.

Implementing a model often faces hurdles, as data science teams typically handle core tasks, but full deployment needs a solid web interface and collaboration with DevOps, data engineers, and backend engineers to manage live data streams. Metal Minds provides a full-cycle team covering all development stages: mathematical core, front-end, back-end, and DevOps. 

When it comes to adoption, even the premier tool will be ineffective if it is no longer used after six months. The main challenge is ensuring the model remains in use after control is handed over. Metal Minds effectively addresses these adoption challenges with our clients.

9. How can investors benefit from implementing advanced AI technologies in metallurgy? What is most attractive about such projects for them?

The formula is simple: the impact of implementing our solutions far exceeds the cost of Metal Minds’ services. This impact is unique and significantly improves existing processes. 

Investors typically care about the speed of change implementation — our projects typically take about 6 months. Unlike capital-intensive initiatives, which can take 1-3 years and have a payback period of 5-10 years, AI initiatives deliver results immediately and pay off within the first year.

10. How do you see the development of AI in metallurgy and other industries over the next 5-10 years?

I see growing interest in AI in heavy industry, especially after the widespread access to ChatGPT. People at all levels have started believing that AI is a real, working, and powerful technology. Engineers are increasingly proposing: “Why don’t we try solving this problem with AI?” While the solution may not actually require AI, the interest in the technology remains strong. This has significantly accelerated the process of initiating new ideas in the workplace. 

Management in heavy industry has traditionally been conservative toward digital technologies, but I notice that this conservatism is weakening, and more AI projects and experiments will emerge.

11. What is your goal for the coming years? What ambitions do you have in the field of technology for the metallurgy industry?

My goal is to create my own factory that is fully autonomous. This factory will not have any employees, ensuring very high safety standards, and all decisions will be made automatically, optimally, and profitably for investors and the environment. Right now, Metal Minds is focused on accumulating successful cases of decision-making optimization. Our goal in the coming years is to fully automate the process of making routine decisions.

12. In your interviews, you often warn about the dangers of uncontrolled AI development. How do you address this?

With the rise of AI, there has been a trend of blindly trusting algorithmic results, especially after the appearance of ChatGPT, where some began to see it as a source of absolute truth. It’s important to set technological guardrails for each algorithm to prevent accidents or harm to people. We should be careful and mindful.

13. What role should ethics play in AI development, especially in such critical industries as metallurgy?

Our data and processes are not related to people or their personal data, so we haven’t encountered ethical issues yet. However, I believe that environmental awareness is much more important in our case — many of our models specifically optimize environmental impact.

14. What advice would you give to companies that are just starting to implement AI and machine learning technologies in their production processes?

Support your workforce by offering training, providing the right tools, and considering financial incentives. Some of the most effective projects come from within the organization. Interest in AI in industry is expected to grow, and those who can adapt to new conditions may find AI becoming a valuable asset. 

Published by Liana P.