Shobhit Kamal-Most Innovative Tech Leaders from USA 2026

Manufacturing has become remarkably good at automating physical work, yet some of its hardest problems still demand human experience, judgment, and time. Shobhit Kamal saw this gap firsthand during his five years at Rivian, where he progressed through manufacturing engineering roles to lead battery manufacturing engineering. Working on complex production systems gave him a close view of the challenges that remain when machines execute tasks efficiently but diagnosing problems and making sense of operational complexity remain difficult. Today, as Founder and CEO of a stealth-mode startup in the industrial AI space, he is working to bridge that gap by exploring how AI can help factories become more intelligent and adaptive. His perspective is further shaped by his role on the board of KKS Group, his family’s diversified business group in India.

In this exclusive conversation with TradeFlock, Kamal discusses practical AI adoption, technology leadership, and building solutions around real industrial needs.

What did your experience at Rivian teach you about solving complex problems and taking risks?

I was one of Rivian’s earliest engineers, where I spent five years developing and scaling battery manufacturing for some of the most complex products in modern engineering. I launched and ramped several automated production lines supporting Rivian’s electric SUVs and trucks.

Manufacturing at this scale is incredibly difficult. A single production line can span a football field, with hundreds of robots, thousands of sensors, and millions of interconnected decisions happening every hour.

During the R2 launch, I recognized a fundamental gap. We had made enormous progress in automating physical work, but not in making factories intelligent. Machines could execute precisely defined instructions, yet diagnosing why something went wrong still depended heavily on human interpretation and experience. Two engineers could investigate the same failure and reach different conclusions.

That realization changed my trajectory. I began thinking about what it would take to give factories the ability to understand their own behavior, learn from mistakes, and help engineers solve problems faster.

Leaving a career I had spent five years building to pursue that problem was a calculated risk. I had seen the problem firsthand, recognized a gap in the industry, and believed the technology had reached an inflection point. Sometimes the biggest risk is continuing to work around a problem you know should be solved differently.

How do you evaluate industrial AI opportunities, and when would you abandon a promising idea?

I use three filters: pain, feasibility, and economic value.

First, is the problem expensive and frequent enough that a customer genuinely needs it solved? Second, can we solve it reliably with the data, infrastructure, and constraints that exist in a real factory? Third, does the improvement justify the cost and effort of implementation?

Industrial environments punish technology that looks impressive in a demonstration but fails under real operating conditions. A technically sophisticated system is irrelevant if deployment takes months, requires significant capital expenditure, or cannot earn the trust of the people using it.

I am willing to abandon an idea if the economics do not work, the data cannot support reliable outcomes, or the implementation burden outweighs the benefit. Being an engineer means becoming attached to solving problems, not to a particular solution.

As AI agents take on industrial decisions, where should human oversight begin?

Autonomy should be earned, not assumed.

AI can significantly amplify intelligence orchestration in manufacturing facilities. But decisions affecting safety, equipment integrity, product quality, or critical process controls require appropriate evaluations and guardrails.

I believe in progressive autonomy: start with recommendations, validate performance, establish clear operating boundaries, and expand authority only when the evidence supports it.

Manufacturers must never overlook the consequences of a confident but incorrect decision. In a factory, errors are physical and the stakes are high. There is no room for mistakes. Human oversight, cybersecurity, and fail-safe mechanisms are fundamental engineering requirements, not optional features.

Where can AI outperform conventional automation in manufacturing, and how do you measure its impact?

I start with the factory floor, not the technology. The question is not where AI can be applied, but where people are spending disproportionate time navigating complexity.

Legacy automation excels at repeatable, predefined tasks. AI becomes valuable when situations are ambiguous: a machine behaves unexpectedly, several process variables interact, or the evidence needed to diagnose a failure is scattered across systems and documentation.

These are problems where experienced engineers often become the bottleneck because their knowledge is difficult to codify and scale.

We’ve seen overwhelming interest from manufacturers and investors alike, reinforcing something I strongly believe: manufacturing is becoming exciting again, and the reindustrialization thesis is real. But excitement alone doesn’t create value. We measure success through tangible operational improvements, including faster troubleshooting, reduced downtime, higher throughput, and better quality.

Ultimately, AI’s impact on manufacturing will not be measured by how sophisticated the technology is, but by whether it can adapt and get better with every production cycle.

What do AI teams underestimate when introducing technology into factories?

They underestimate the gap between a technology working in isolation and working in production.

Factories are not clean software environments. Equipment spans generations, data is fragmented, documentation is incomplete, and every production interruption has a real cost. Operators and engineers also have established workflows that exist for good reasons.

Companies should begin with a narrowly defined problem, validate against historical operating conditions, and involve frontline engineers from the outset. Solutions must fit existing processes rather than assume factories will reorganize themselves around new technology.

Trust is earned through evidence, repeatability, and results, not through a compelling AI demo.

What principle guides your approach to innovation and long-term responsibility?

As a second-generation founder, I grew up watching my dad build and scale multiple enterprises from scratch. He instilled in me a belief that value creation and continuous innovation must sit at the heart of any enterprise. Growth alone is not the objective. Building something that solves meaningful problems, creates lasting value, and stands the test of time is what matters.

My experience in manufacturing reinforced that philosophy. Behind every machine, process, and production target are people responsible for keeping operations running. Technology should make them better at what they do, make expertise more accessible, and amplify holistic intelligence.

As a founder, I apply the same principle: start with a real problem, challenge conventional approaches, and measure success by the value created, not the technology deployed. Ultimately, innovation matters only when it makes a meaningful impact and elevates society.