From Liquid-Cooled PCs to Frontier AI: Why I Build the Way I Do
I have been building since 2006: liquid-cooled machines as a teenager, industrial robotics, real estate media, logistics, and now an AI lab. The through-line is simple: build things that hold up under load.
2006: where it started
I got my first real machine in 2006, when I was ten, in North India. Personal computing was young there and information was scarce, so I learned by taking things apart. At fourteen I built one of the first liquid-cooled systems in the region, sourcing parts from the United States because at that time only large corporations like Infosys ran liquid-cooled machines in India. It worked. It also taught me the lesson that still drives everything I build: thermals, tolerances, and failure modes matter more than specs on a box.
What speed skating taught me about engineering
From 2007 to 2014 I sprint skated for India: 300m Time Trial, 500m and 1000m sprints, over 75 gold medals at district, state, and national levels. At the 2011 international meet in Namwon, South Korea, I finished fifth in the world in the 300m Time Trial.
Sprint skating is a hundredth-of-a-second sport. You do not get style points for almost winning. That mindset transferred directly to engineering: measure everything, remove what does not work, and respect the clock.
The pivot years
In 2014 I moved to Canada and studied Electromechanical Engineering, Robotics and Automation at Centennial College in Toronto: ABB robot cells, PLC programming, hydraulics, pneumatics, machine design. Before that, a full-ride scholarship took me to Aichi Shukutoku University in Nagoya, and time spent in Japanese foundries with my father, who built sand plants for factories across India, taught me how serious manufacturing actually works.
Then came Zebra Robotics, where I joined as a Python instructor in 2017 and left as Business Development Manager after franchising the first location to Cary, North Carolina. Then British Columbia: two years growing revenue at Pallet Terminal, while founding Dipole Media Production, a real estate media company that crossed $100,000 ARR in its second year on the strength of accurate floor plans and listing media.
What all of this has to do with Dipole ML
In 2024 Dipole Media became dipole.studio, focused on AI workflows and AEO-driven growth for local businesses. In the same year I came back to Pallet Terminal, this time as a one-third owner, and turnover has tripled since.
Dipole ML is the sum of all of it. A model company run by someone who has cooled computers by hand, wired PLCs, run a warehouse, and scaled a services business. We ship Raven 1.0 on October 1, 2026, and we will hold it to the only standard I have ever respected: does it perform under real load?
About Saurabh Sharma
Founder & CEO of Dipole ML, launching Raven 1.0 on October 1, 2026. Two decades across computing hardware, industrial robotics, media, logistics, and AI.
