Atomic Industries CEO Aaron Slodov demos AI-driven mold manufacturing from Detroit factory floor

Summary

Aaron Slodov, CEO of Atomic Industries, joined live from his Detroit factory floor to explain what the company is building: software and AI systems that can engineer injection molds in minutes rather than the weeks or months it currently takes skilled tradespeople.

The core problem Atomic is solving is that mold-making — a prerequisite for mass-producing almost anything in plastic or composite — is deeply dependent on tribal knowledge that takes years, sometimes decades, to develop. Slodov's bet is that you can capture that expertise, encode it into a model, and then use the factory itself to close the feedback loop. The technical architecture he describes is closer to how early self-driving systems were built than to a pure LLM play: numerical physics solvers characterize the problem first, then models trained on top of that simulation data learn to predict mold geometry and cooling channel placement quickly. The factory generates real-world data that improves prediction accuracy over time. LLMs, in his framing, are useful for coordination and factory operations — not for the physics.

On the floor during the call, Atomic's machines were cutting molds for a drone program and running production work for the New York City Subway Transit Authority.

The China comparison

Slodov's competitive framing is pointed. American high-mix manufacturers, he argues, routinely price themselves out of jobs before they even engage a customer — spending time evaluating whether margin is good enough rather than quoting instantly. Chinese competitors respond fast, price aggressively, and absorb complexity. Atomic's answer is software that can generate instant quotes across manufacturing processes, compressing the sales cycle and removing the gatekeeping that has pushed US shops toward only the highest-margin work.

His broader argument is that the US lost manufacturing depth gradually, and the consequence isn't just cost structure — it's the erosion of the institutional knowledge needed to build anything complex. Training a person to this level takes roughly a decade. Transferring that knowledge into a machine and making existing workers 100 times more productive is, in his view, the only realistic path to rebuilding it fast enough to matter.

Talent pipeline

On attracting the next generation to factory work, Slodov argues the lever is twofold: financial incentive and intellectual challenge. If a factory worker in an AI-optimized shop is productive enough, there's no structural reason they can't earn what a senior engineer at a large tech company earns. He also argues that more prominent industrialists need to actively promote the sector, and that factory tours and direct exposure matter more than curriculum changes alone.

Humanoid robots

Slodov says humanoids could eventually be useful for lower-precision tasks around the factory floor, but current dexterity limitations make them marginal for the precision work Atomic does. He isn't betting on them near-term.