The engineer teaching AI the difference between real and fake
Tech Funding News Abhinaya Prabhu
A Delhi founder in Vancouver now checks which human-task videos are real for robots to learn from. It’s a job built for the age of fake everything.
Based on reporting by Tech Funding News, Abhinaya Prabhu — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Saurab Dhir spends his workdays at Mecka AI judging whether recordings of people doing physical tasks are authentic. The clips come from contributors around the world, and the point is blunt: robots will learn from them later, so the company needs to know the footage is real before it goes anywhere near a model.
That makes Dhir part technician, part skeptic. He built the systems that help verify authenticity, which also means he lives close to the many ways people try to fake it. In a moment when synthetic text, images, and video have made “is this real” a daily problem, he is one of the few people paid to answer it with instruments instead of gut feeling.
The temperament seems older than the job. Dhir grew up on the edge of Delhi, the son of a father who ran a manufacturing business and a mother who kept the house. As a kid, he pulled an RC helicopter apart and never rebuilt it, later bricked a phone trying to jailbreak it. For his 14th birthday, his father made him build his own gaming computer by doing the sourcing, research, bargaining, and buying himself. That mix of curiosity and hard-nosed diligence followed him into engineering.
His family also shaped the moral side of the work. His mother taught empathy, his father showed him that honest work can cost less than dishonest work. Dhir carries that into how he thinks about contributors: like a fair referee, he watches for cheating but won’t call it without proof. He says the line between a cheat and a perfectionist can be only a few frames of hesitation.
The path to Canada was less polished. He came to Simon Fraser after seeing photos of the campus online and imagining a bigger life, then found his first month in the country to be the worst stretch he’d had. He slept on a bare mattress under a borrowed blanket, with spiders on the wall, before moving onto campus and later into his own place on Burnaby Mountain. A call to Rohan Parmar helped get him through that first hard night.
Outside the lab, Dhir keeps two constants: guitar and chai. He has played since he was 13, with Metallica, Lamb of God, Megadeth, and Iron Maiden in the mix, and he spent years in one apartment making fresh chai every morning for his co-founder Mikhail Sinitcyn. It’s a tidy fit, really. Metal and bouldering both reward precision, patience, and the ability to commit after a lot of failed attempts.
My take — AI-written commentary, not fact-checked reporting
This is the unglamorous side of AI that actually matters: not bigger demos, but better gatekeepers. Everyone wants to build the machine; fewer want to do the dreary work of checking whether its training diet is garbage. That’s where the real leverage is, and also where the least applause lives.
Read more about this at: Tech Funding News