He taught himself modern AI for six months before writing a line of code. Then he built a company that teaches machines to do the serious stuff.
What makes an AI agent unique is its ability to autonomously carry out a task while dynamically reacting to information.
Long before "AI agent" was a pitch deck phrase, Asver was chasing a stubborn idea: catch the world in real time. In 2008 he co-founded Scoopler, a real-time search engine backed by Y Combinator and angels including Ron Conway. It pivoted into JustSpotted, a site that tracked the live whereabouts of celebrities from public signals - part stalker-fiction, part early data play.
In 2011, Google scooped up the team to work on Google+. Asver stayed on the inside for four years - shipping on Google Photos, then as a hardware product manager on Nexus and Android. It is a useful detail: he has built both atoms and bits, both consumer and infrastructure.
Then came a left turn. From 2015 to 2017 he produced music and DJ'd as AJX. He calls himself, with a straight face, a "retired DJ." The instinct - read the room, react in real time, keep the floor moving - is not unrelated to the agents he builds today.
Asver landed in fintech as it was getting interesting. At Coinbase he moved through consumer product into the deep end of payments risk and data, eventually leading Data & Payments Risk. At Brex he became Director of Product for the Platform, sitting close to the operational and compliance bottlenecks that quietly throttle every fast-growing financial company.
That is where the Parcha thesis was forged - and where he met his co-founder, Miguel Rios-Berrios. Both had watched smart people drown in manual reviews. Both suspected software could lift the load. Investors at Kindred Ventures and Initialized Capital noticed something else: a founder who is genuinely technical and can sell, who ideates and prototypes on several threads at once. Initialized's Brett Gibson described them working "crazy fast."
Parcha's core value is "Believe." Yes, it is borrowed from Ted Lasso. No, there is no backup plan.
Before writing any Parcha code, Asver spent six months interviewing AI researchers and publishing a Substack, "A Hitchhiker's Guide to AI." He treated learning as the first product.
He hires curious generalist engineers and teaches them AI, arguing you don't need years of ML experience to ship production AI - only willingness to experiment.
Parcha runs on a mix of Grep research agents, "OpenClaw" ops agents and coding agents working alongside human engineers. The company eats its own dog food.
The endgame Asver describes is not one super-model but a faculty: "hundreds, eventually thousands of experts, taught by human domain specialists who can teach Grep what good work looks like." Apprenticeship, scaled. The expert as a thing you can summon.