Building software has long followed a reliable playbook: Write code, test, deploy and iterate. But when deterministic algorithms give way to artificial intelligence models that learn and adapt, the old playbook falls apart. AI development infrastructure has become essential for enterprises trying to move projects from prototype to production.

The shift stems from a fundamental difference in how AI products work. Traditional software follows predictable paths, where every decision traces back to human-written code. AI introduces uncertainty, with models producing outputs based on probabilities rather than fixed rules. That transforms development into something closer to research, requiring teams to run hundreds of experiments before finding approaches worth deploying, according to Ketan Umare (pictured), co-founder and chief executive officer of Union.ai.

“Research becomes a part of the software development process, and we are not used to doing research in a software community,” Umare said. “People who are used to doing research are people who are building drugs. If you go and talk to the CEO … they don’t say, ‘Make every drug experiment I do, take it to production.’ They say, ‘Reduce the cost of my drug experimentation.'”

Umare spoke with theCUBE’s Dave Vellante at AWS re:Invent, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how Union.ai’s multicloud orchestration platform helps enterprises manage AI development’s experimental nature while maintaining security and controlling costs.

AI development infrastructure connects and secures multicloud workflows

Union.ai addresses complexity through an architecture separating control from execution. The company runs its control panel on Amazon Web Services Inc. while connecting to customer environments wherever data resides. Organizations keep sensitive information in their own clouds or on-premises while benefitting from unified orchestration, according to Umare.

“That’s where your data resides, data planes run, like your compute, your inferencing, your training cycles all run in the cloud, and we can seamlessly connect these multiple clouds into one common way,” he said. “I’ve gotten to a point where I actually get Claude code to write some of my examples that run on the system, and then, in one shot, it’s able to run. It’s literally YOLOing onto the platform because it can build infrastructure as it needs dynamically, and then pre-provision it once it’s done. That’s the power of this AI development infrastructure that we’re building.”

The platform handles containerization, data movement, security permission and graphics processing unit scheduling. Developers write standard Python locally, and with a single action, the system containerizes their work and deploys it. Union.ai targets one-second deployment from local development to remote executive, according to Umare.

“If you go and talk to an AI engineer and say, ‘By the way, you take your agent to production, now first write Terraform, and then go deploy and then write a CI/CD pipeline,’ they’ll go, ‘Come on. I’m not going to do all of this because I don’t even know if this thing works. It’s too complicated.'”

Cost optimization runs through the design of the AI development infrastructure. When engineers work on project variations, the system caches shared results to avoid duplicate effort. Union.ai uses AWS’ reserved instances to manage complex pools, queueing jobs when resources are constrained. The 46-person company has quadrupled revenue this year while keeping costs lean, according to Umare.

“I think the important thing to understand is there’s no way out of doing research, because this is a research-based product; it’s non-deterministic,” he said. “But you can reduce the cost  by doing smart things.”

Here’s the complete video interview, part of theCUBE’s coverage of AWS re:Invent:

Photo: SiliconANGLE