Keven
Wang

There is a particular kind of defect - a hairline crack in a lithium-ion battery cell, a microscopic void on a connector pin, a surface pit barely 40 microns wide on an automotive part - that a human inspector will miss on the 200th look of a long shift but that will cascade into a recall costing tens of millions. Keven Wang has built his career on that gap.

Wang is the Co-Founder and CEO of UnitX, a Santa Clara-based AI robotics company rewriting the quality control playbook for the factories that make EVs, batteries, and electronics. His systems do not merely detect defects. They do so at production speed, with 9x fewer false passes than a trained human inspector, and they can be taught a new defect category using three sample images and a generative AI pipeline that Wang's team built from scratch.

The company has raised $92 million in total funding, runs 820+ systems on production lines in 135+ factories across the globe, and inspects $6.1 billion in manufactured products every year. The top-two EV manufacturers in the world are customers. So are the top-ten automotive Tier 1 suppliers.

The path from Keven Wang's University of Michigan engineering degree to running a robotics company is less surprising than it looks. He graduated in 2013 with a 3.71 GPA in Computer Science and joined SalesforceIQ - the CRM startup Salesforce would acquire - where he shipped more than 25 product features in three years, including the Chrome Extension and Outlook Add-in that hundreds of thousands of sales professionals used daily. Fast iteration, customer obsession, measurable impact. These habits are now in UnitX's DNA.

In 2014, he enrolled in Stanford's Master of Science program in Computer Science with a concentration in artificial intelligence, finishing in 2018. That same year, with co-founders Adam Yang and Weixiong Zheng - engineers with roots spanning Stanford, MIT, and Google - Wang founded UnitX. The founding thesis was precise: deep learning had crossed a threshold where it could outperform human vision at repetitive inspection tasks, but nobody had packaged it into something a factory operations team could actually deploy and maintain.

What separated Wang from the parade of machine-vision startups that came before UnitX was the commitment to understanding the physical world on its own terms. He personally visited more than 135 manufacturing facilities worldwide - walking production lines, talking to quality engineers, watching where human inspectors hesitate - before deciding what product to build.

Wang insists UnitX is not a software company. He calls it a Robotics 2.0 company - one that integrates proprietary AI models, a patented software-defined imaging system, and edge compute hardware into a single deployable unit. That distinction matters on the factory floor. A pure-software approach inherits whatever optics the customer already owns. UnitX's OptiX system can generate up to 2³² distinct lighting patterns, adapting dynamically to different part geometries and surface textures.

The company's latest platform, FleX, launched in December 2025, and its performance numbers are striking: 9x lower escape rates on subtle defects compared to traditional methods, deployment in under one week, and a return on investment timeline under 12 months with an average of $1.3 million returned per production line.

The technical moat, though, lives in GenX. Training AI inspection models historically required hundreds or thousands of defect images - images that can take months to collect on low-defect-rate production lines. GenX creates photorealistic synthetic defect images from as few as three real samples, using a generative AI pipeline that Wang's team developed specifically for industrial vision. A leading EV battery supplier using GenX cut model training time by 80%.

Spend any time reading Wang's interviews and one observation keeps surfacing: the thing blocking AI adoption in factories is not the AI. It is the people. "The greatest challenge in AI adoption isn't technical but human," he has said - specifically: aligning teams, establishing shared goals, and helping production staff become comfortable with a system that occasionally disagrees with their seasoned judgment.

Wang brings something unusual to that conversation: three languages (English, Chinese, and German) and the personal credibility of having stood on 135 production floors. He is not pitching AI in the abstract. He is talking to a quality supervisor in Shenzhen, a battery engineer in Stuttgart, and a Tier 1 supplier in Detroit, and he has walked their lines.

His philosophy - "automate everything and maximize human intellect" - is deliberately not "replace the humans." The framing is that repetitive visual inspection is a poor use of human attention. Free that attention and you get better engineers focusing on harder problems.

For operators, Wang recommends deploying AI when production reaches sufficient volume with dedicated lines, and validating with representative samples rather than exhaustive testing. Prove ROI fast. Productize serviceability from day one. The advice reads like someone who has seen too many pilots stall.