MayimFlow wants to stop data center leaks before they happen

The mania astir information centers has a batch of companies looking to go “picks and shovels” providers, by gathering profitable businesses that are ancillary to the golden unreserved gigs of selling server entree oregon grooming starring AI models.

MayimFlow, the Built World signifier victor astatine this year’s TechCrunch Disrupt, is simply a bully example. The startup is focused connected fundamentally 1 task: preventing damaging h2o leaks.

Data centers usage a batch of water, and that h2o tin contiguous a large risk, adjacent if a leak is small. Founder John Khazraee told TechCrunch that galore information centers lone person reactive solutions for h2o leaks. That tin saddle companies with downtime and acceptable them backmost millions of dollars if 1 occurs.

Khazraee would know. He spent much than 15 years gathering infrastructure for IBM, Oracle, and Microsoft. With MayimFlow, he’s developed a operation of IoT sensors and edge-deployed instrumentality learning models that tin observe signs of impending leaks.

“I’ve noticed these issues successful information centers, and the lone solution they had was: ‘when the leak happens, we find out,’” helium said successful an interview. “Now you person to walk a batch of wealth to spell remediate the situation. Now you got to crook disconnected the servers. Now the information is being disrupted. So I decided to bash thing astir it.”

Khazraee assembled a tiny squad with plentifulness of acquisition to instrumentality connected this challenge. Jim Wong, MayimFlow’s main strategy officer, has spent decades moving with information centers. Chief exertion serviceman Ray Lok has carved retired a vocation successful h2o absorption and IoT infrastructure.

Beyond the scariest consequences of unnoticed h2o leaks, there’s an constituent of frugality to what Khazraee is trying to bash with MayimFlow, which helium said comes from his childhood.

“I grew up successful a very, you know, I don’t privation to accidental the connection ‘poor’ family, but we weren’t the astir well-off family,” helium said. “And my dada would always, erstwhile I was taking a shower, archer me, ‘Hey, you’re successful the ablution excessively long. Are you singing successful there?’”

As helium grew up, Khazraee said helium recovered himself ever reasoning astir however to marque things much businesslike — particularly arsenic helium studied to go an engineer. In college, helium worked astatine a installation that collected frying lipid from restaurants and converted it to biodiesel. It was a “messy job,” helium said, but helium liked the end result.

Khazraee is present trying to premix this penchant for frugality with the acquisition of his team. He believes they tin springiness information halfway operators 24 to 48 hours of precocious informing that repairs volition beryllium needed.

MayimFlow has collected a glut of illustration information from assorted concern h2o systems that let them to marque these kinds of predictions, helium said. The institution tin supply sensors that tin assistance show a h2o strategy oregon plug its instrumentality learning models into existing ones if companies already person akin hardware.

Khazraee said helium wants to bring this solution beyond information centers to commercialized buildings, hospitals, manufacturing facilities, and perchance adjacent utilities. In his eyes, immoderate institution that wants to spot leaks early, oregon optimize h2o usage, could beryllium a customer.

Khazraee thinks it’s specified a large accidental he’s turned down roles astatine aggregate large tech companies portion gathering MayimFlow implicit the past 2 years.

“I truly judge successful the vision. I judge successful the interaction that we’re making,” helium said. “Water is becoming 1 of those large issues successful our world.”