Suman
Kanuganti

There's a company in San Francisco where 53 people are quietly building something that sounds obvious once you hear it: what if your AI remembered everything you knew, forever? That company is Personal AI. The person running it is Suman Kanuganti - and he has been thinking about this longer than most people have been using ChatGPT.

Kanuganti's path to this idea wasn't straight. He came from India, trained as an electrical engineer at Kakatiya University, sharpened his technical edge in robotics at the University of Missouri-Columbia, and then - unusually - added an MBA from UC San Diego's Rady School of Management. At Qualcomm and Intuit, he learned what enterprise-grade systems actually look like under pressure. Both companies. Real scale.

In 2015, he co-founded Aira - long before AI was everyone's favorite word. The premise: use AI and augmented reality glasses to give blind and low-vision people access to visual information, anywhere, anytime. The New York Times called it a "godsend." TIME Magazine put it on their Best Inventions list. Fast Company named it a World Changing Idea. Kanuganti built Aira to $50M in revenue before stepping back.

He would later describe those early years as "physical AI when it was mostly too early." He was a decade early to the wave. He built anyway.

In 2020, while the world was just starting to understand what language models could do, Kanuganti started Personal AI. The initial concept: a digital version of you, built from your own data, facts, and opinions. Not a chatbot. Not a search engine. A memory extension - personal, persistent, and private.

When OpenAI launched ChatGPT in late 2022, Suman Kanuganti had a problem that most founders would envy: everyone suddenly wanted what he had been building for two years. The problem was that the market couldn't differentiate - "confusing to investors, confusing to customers," as he put it. His consumer-focused product was getting lost in the noise of every AI startup suddenly claiming the same territory.

So he pivoted. Not the product - the customer. He went enterprise. Specifically, he went after companies with highly sensitive data, the ones that couldn't just send everything to OpenAI's servers and hope for the best. Healthcare. Legal. Finance. Organizations where data sovereignty is existential.

The enterprise response was immediate: "Where have you been?" Personal AI had the private, secure, memory-first architecture that large organizations needed. The pivot worked.

Today, Personal AI's platform is trusted by Microsoft, NVIDIA, HPE, Verizon, AT&T, and T-Mobile. NVIDIA named the company its top independent software vendor for telco. The pitch: Personal Language Models that deliver 10-30x better cost efficiency than general-purpose large language models - because they're trained on specific knowledge, not the entire internet.

Kanuganti's stated vision isn't incremental. He thinks public companies will, in the not-so-distant future, report their workforce composition with a line item for AI workers - something like "10% AI workforce powered by personal AI." He's building the platform to make that accounting real and auditable.

The architecture is built on five memory primitives: encoding, stabilizing, storing, retrieving, and updating. It's not retrieval-augmented generation bolted onto a general model. It's a persistent AI identity - one that grows more personal through accumulated experience, transfers across systems, and operates under centralized compliance governance.

The use case that captures the imagination: what happens to institutional knowledge when an expert retires? With Personal AI's memory platform, that knowledge doesn't walk out the door. It stays - searchable, queryable, deployable as an AI that can answer "What would [Name] do?" with real fidelity to how that person actually thought.

That essay - about grief, memory, and technology colliding - reveals something about the founder that the pitch decks don't. Kanuganti thinks about AI as a human problem, not just an engineering one. As an AI developer and father (his words), he wrote an open letter to students about AI in 2023. He tweeted about his 6-year-old's thoughts on the future of AI.

He calls the current AI landscape an "attention economy" problem, and argues for what he calls a "connection economy" - AI that deepens relationships and extends human capability rather than capturing eyeballs for advertisers.

There's a pattern in how Kanuganti operates. At Qualcomm in 2007, he was a systems engineer learning what enterprise-grade looks like. At Intuit from 2008 to 2015, he spent seven years as an engineering manager - patient, methodical, watching product development at scale. Then he left to found Aira when "physical AI" was mostly a research project.

He's been accused - approvingly - of being perpetually early. The accessibility wave. The personal AI wave. The enterprise AI memory wave. Each time, the market catches up and finds him already there, with infrastructure built.

His mistake, he says candidly, was staying in San Diego too long. "You need to be part of the conversation," he told a podcast host about relocating to San Francisco for fundraising. He moved. The conversations changed. The capital followed.

He's spoken at TechCrunch Disrupt, CES, SXSW, and the Qualcomm Snapdragon Summit. He's logged 100 podcast appearances - a running tally he hasn't slowed down on. The through-line isn't self-promotion. It's a belief that the idea won't spread unless he personally carries it room to room.

He has 10 patents. He holds an MBA and two graduate degrees, one in robotics, one in business. His written voice on Medium is direct, personal, occasionally philosophical - a CEO who writes about his kids' questions and about deceased colleagues he wishes could still answer Slack messages. Someone who takes seriously the gap between AI as marketed and AI as actually experienced by people.

That gap is what he built his career trying to close.