AI agents are all the rage, a trend driven by the generative AI and large language model (LLM) boom these past few years. Getting people to agree on what exactly AI agents are is a challenge, but most contend they are software programs that can be assigned tasks and given decisions to make — with varying degrees of autonomy.
In short, AI agents go beyond what a mere chatbot can do: they help people get things done.
It’s still early days, but the likes of Salesforce and Google are already investing heavily in AI agents. Amazon CEO Andy Jassy recently hinted at a more “agentic” Alexa in the future, one that’s as much about action as it is words.
In tandem, startups are also raising cash off the hype. The latest of these is German company Juna.ai, which wants to help factories be more efficient by automating complex industrial processes to “maximize production throughput, increase energy efficiency and reduce overall emissions.”
And to pull that off, the Berlin-based startup today said that it has raised $7.5 million in a seed round from Silicon Valley venture capital firm Kleiner Perkins, Sweden-based Norrsken VC, and Kleiner Perkins’ chairman John Doerr.
Self-learning is the way
Founded in 2023, Juna.ai is the handiwork of Matthias Auf der Mauer (pictured above, on the left) and Christian Hardenberg (pictured above, right). Der Mauer previously founded a predictive machine maintenance startup called AiSight and sold it to Swiss smart sensor company Sensirion in 2021, while Hardernberg was the former chief technology officer at European food delivery giant Delivery Hero.
At its core, Juna.ai wants to help manufacturing facilities transform into smarter, self-learning systems that can deliver better margins and, ultimately, a lower carbon footprint. The company focuses on so-called “heavy industries,” — industries such as steel, cement, paper, chemicals, wood and textile with large-scale production processes that consume lots of raw materials.
“We work with very process-driven industries, and it mostly involves use-cases that use a lot of energy,” der Mauer told TechCrunch. “So, for example, chemical reactors that use a lot of heat in order to produce something.”
Juna.ai’s software integrates with manufacturers’ production tools, like industrial software from Aveva or SAP, and looks at all its historical data garnered from machine sensors. This might involve temperate, pressure, velocity, and all the measurements of the given output, such as quality, thickness and color.
Using this information, Juna.ai helps companies train their in-house agents to figure out the optimal settings for machinery, giving operators real-time data and guidance to ensure everything is running at peak efficiency with minimal waste.
For example, a chemical plant that produces a special kind of carbon might use a reactor to mix different oils together and put it through an energy-intensive combustion process. To maximize the output and minimize residual waste, conditions need to be optimal, including the levels of gases and oils used, and the temperature applied to the process. Using historical data to establish the ideal settings and taking real-time conditions into account, Juna.ai’s agents supposedly tell the operator what changes they should be making to achieve the best output.
If Juna.ai can help companies fine-tune their production equipment, they can improve their throughput while reducing energy consumption. It’s a win-win, both for the customer’s bottom line and its carbon footprint.
Juna.ai says it has built its own custom AI models, using open-source tools such as TensorFlow and PyTorch. And to train its models, Juna.ai is using reinforcement learning, a subset of machine learning (ML) that involves a model learning through its interactions with its environment — it tries different actions, observes what happens, and improves.
“The interesting thing about reinforcement learning is that it’s something that can take actions,” Hardenberg told TechCrunch. “Typical models only do predictions, or maybe generate something. But they can’t control.”
Much of what Juna.ai is doing at present is more akin to a “copilot” — it serves up a screen that tells the operator what tweaks they should be making to the controls. However, many industrial processes are incredibly repetitive, which is why enabling a system to take actual actions is helpful. A cooling system, for instance, might require constant fine-tuning to ensure a machine maintains the right temperature.
Factories are already well accustomed to automating system controls using PID and MPC controllers, so this is something that Juna.ai could feasibly do, too. Still, for a fledgling AI startup, it’s easier to sell a copilot — it’s baby steps for now.
“It’s technically possible for us to let it run autonomously right now; we would just need to implement the connection. But in the end, it’s really all about building trust with the customer,” der Mauer said.
Hardenberg added that the benefit of the startup’s platform doesn’t lie in saving labor, noting that factories are already “quite efficient” in terms of automating manual processes. It’s all about optimizing those processes to cut costly waste.
“There’s not a lot to gain by removing one person, compared to a process that costs you $20 million in energy,” he said. “So the real gain is, can we go from $20 million in energy to $18 million or $17 million?”
Pre-trained agents
For now, Juna.ai’s big promise is an AI agent tailored to each customer using their historical data. But in the future, the company plans to offer off-the-shelf “pre-trained” agents that don’t need much in the way of training on a new customer’s data.
“If we build simulations again and again, we get to a place where we can potentially have simulation templates that can be reused,” der Mauer said.
So if two companies use the same kind of chemical reactor, for instance, it might be possible to lift-and-shift AI agents between customers. One model for one machine, is the general gist.
However, there’s no ignoring the fact that enterprises have been hesitant to dive head-first into the burgeoning AI revolution due to data privacy concerns. These concerns are lost on Juna.ai, but Hardenberg said that it hasn’t been a major issue so far, partly due to its data residency controls, and partly due to the promise it gives customers in terms of unlocking latent value from vast banks of data.
“I was seeing that as a potential problem, but so far, it hasn’t been such a big problem because we leave all data in Germany for our German customers,” Hardenberg said. “They get their own server set up, and we have top-notch security guarantees. From their side, they have all this data lying around, but they haven’t been so effective at creating value from it; it was mostly used for alerting, or maybe some manual analytics. But our view is that we can do much more with this data — build an intelligent factory, and become the brain of that factory based on the data they have.”
A little more than a year since its foundation, Juna.ai has a handful of customers already, though der Mauer said he’s not at liberty to reveal any specific names yet. They are all based in Germany, though, and they all either have subsidiaries elsewhere, or are subsidiaries of companies based elsewhere.
“We’re planning to grow with them — it’s a very good way to expand with your customers,” Hardenberg added.
With the fresh $7.5 million in the bank, Juna.ai is now well-financed to expand beyond its current headcount of six, with plans to double-down on its technical expertise.
“It’s a software company at the end of the day, and that basically means people,” Hardenberg said.
