Aravind Bala did not leave Microsoft with a revelation. He left with a colleague, a tolerance for uncertainty and the conviction that waiting for the immaculate startup idea was another way to remain comfortably employed. After roughly 16 years inside Microsoft, most recently as a partner engineering manager, he and Anoop Gupta resigned in 2015. They had worked together on Office Mix. They knew they wanted to build. The troublesome detail was what.
There is a tidier version of entrepreneurship in which the founder spots an obvious injustice, sketches the answer on a napkin and is vindicated after a tasteful montage. Bala's version has more wrong turns and is therefore more useful. He and Gupta began with Nextio, a professional messaging service built around a clever economic complaint: people receive valuable solicitations and hopeless spam, but the platform collecting the money treats both kinds of attention as its own property.
The product would pay recipients to answer. A recruiter with a dream job might be worth hearing; the proverbial prince offering treasure could remain in exile. It was an elegant idea with the grave defect of not becoming a good business. Yet Nextio contained a smaller tool called Career Insights, which mapped possible career moves by analyzing large collections of résumés. Users found it mildly interesting. Recruiters found the machinery behind it much more interesting. They did not particularly want to pay people to read messages. They wanted help finding the right people.
A month to find the signal
By Bala's telling, the pivot was less mystical than mathematical. The team had about six months of money left. It worked backward from survival: how much product could be built, how quickly could it be sold, and what would count as proof? The decisive question was, “What could we build in one month that would actually allow us to get signal and survive?” The answer was a recruiting search product assembled from technology they already had and a customer request they could no longer pretend not to hear.
Bala calls B2B software “kind of like debuggable.” It is a phrase only an engineer could make sound reassuring. Watch what customers do. Listen for the repeated complaint. Isolate the valuable behavior. Change one thing and run the system again. The market is not a machine, of course. It has moods, budgets and vice presidents. But the instinct rescued the company from the founder's most expensive temptation: defending the beauty of the original idea.
Two calculations made the leap possible. One was the regret-minimization idea associated with Jeff Bezos. Bala could imagine tolerating a startup that failed; he was less willing to imagine reaching the end of his career without trying one. The other calculation was about people. He had worked with Gupta for several years and wanted to build with him. A promising market can survive a dreary meeting. A company cannot survive an endless procession of them. Trust in the collaborator mattered before confidence in the concept.
Microsoft had supplied rehearsals. Bala gravitated toward incubation teams and zero-to-one projects, sometimes proposing an idea, securing internal funding, assembling the group and delivering the product. Yet an internal startup still borrowed the larger company's gravity. Customers, dependencies and priorities arrived with invisible strings attached. Leaving turned the simulation into the thing itself. There would be no corporate floor beneath the experiment, only the runway and whatever the team could learn before reaching its end.
“I've always liked building stuff.”Aravind Bala
Building had been the constant long before the company acquired its name. Bala studied computer science at IIT Bombay, then earned a master's degree in the subject at Ohio State. He joined Microsoft's Natural Language Group in 1999, briefly left for advertising analytics startup Paramark in 2000, and returned after the dot-com collapse wiped out the young company's footing. For the next decade and a half, his jobs moved through natural user interfaces, Windows, OfficeLabs, Bing and Office.
The products changed, but one problem kept reappearing: discoverability. Among his Microsoft inventions was a system for searching the commands available inside an application. Later came Bing advertising, relevance and search personalization. How does software infer what a person means? How does it surface the action, result or clue buried beneath too much information? SeekOut eventually aimed that family of questions at careers.
The résumé is only one witness
Recruiters knew the problem. A fine engineer may have a neglected professional profile and an eloquent trail of code. A scientist's work may live in papers and patents. A security-cleared specialist may be visible only through a peculiar combination of experience. Conventional keyword search rewards the person who describes a skill fluently, which is not always the person who practices it well.
SeekOut joined signals from public profiles, GitHub, academic work and patents, then added tools for searching, filtering and engaging candidates. In 2019, it had 12 employees, more than 75 enterprise customers and a new $6 million Series A. By early 2021, the company said annual recurring revenue had grown more than tenfold since that round. A $65 million Series B valued it at roughly $500 million. Nine months later, a $115 million Series C put the valuation at $1.2 billion.
Numbers of that size have a way of making every earlier choice seem inevitable. It was not. Recruiting expanded furiously, then contracted. SeekOut grew and later cut staff in 2023 and 2024. Its product emphasis moved with the market, from broad talent intelligence toward tools that source, screen and engage candidates with AI agents. The current platform searches more than one billion profiles and says it serves over 750 customers. By September 2026, its leadership page named Bala co-founder and CEO.
After skills become cheap
A promotion into the chief executive's chair would be enough plot for most profiles. Bala has chosen a more interesting question. In his 2026 essay “Becoming AI Forward,” he argued that AI alters what companies ought to seek in a person. Technical skills once served as both qualification and convenient proxy. If AI can help a capable worker cross a skill gap on demand, the proxy weakens.
His proposed replacements are more human and far harder to put into a database. “Taste, judgment, and ambition matter far more when AI can bridge skill gaps on demand,” he wrote. Taste decides what is worth making. Judgment distinguishes the plausible answer from the wise one. Ambition determines whether cheaper execution produces a brighter idea or merely a larger heap of features.
This is an awkward thesis for recruiting software, which has prospered by making attributes searchable. Skills can be named. Judgment usually appears only after a consequential decision. Taste is legible in the edit, the refusal and the product someone had the restraint not to ship. Ambition can look identical to noise until time tells them apart. Bala's new problem is therefore the inverse of his old one. Search engineering made scattered credentials visible. AI hiring must help people consider the qualities that resist being converted into credentials.
When execution becomes abundant, choosing well becomes the scarce work.
An engineer with an evening rule
The founder mythology enjoys the all-consuming schedule because exhaustion photographs well. Bala has described startup work as consuming, too, but paired that observation with a domestic rule. During the Nextio years, he tried to reach home by six, spend a few hours with his wife and children, and return to work after they slept. “In a startup, you are always severely under-resourced,” he said. The sensible response was not indiscriminate effort. It was choosing work with the greatest return.
His setup in 2017 had the cheerful excess of an engineer granted purchasing authority: identical home and office computers, two 27-inch monitors at each location, and 64 gigabytes of memory for a custom in-memory database. The office was for the charge of a small team bouncing ideas around and testing restaurants at lunch. Home was for hard problems without interruption. He liked Office partly because, after building command search, he knew enough of its corners to make it do nearly anything.
He also distrusted the comfort of an agreeable feed. Facebook was useful for friends and articles, he said, but its habit of returning more of what one already liked made contrary viewpoints a deliberate errand. The instinct fits the company story. Nextio became SeekOut because the founders allowed contradictory evidence to win.
The useful error
The charming lesson would be that failure is secretly success. Usually it is not. A bad product can remain loyally bad. What made this mistake useful was the discipline applied around it. The team put an adjacent feature in front of people, noticed who became unusually interested, and surrendered the grand idea to the smaller demand. Then it built quickly enough to learn before the money ran out.
Bala's career is full of search systems, but the most important search was conducted by a handful of founders looking for their own company. They found it where good search often finds things: not in the loudest description, but in the neglected evidence nearby. Now the company that learned to see beyond a résumé must learn to see beyond skills. That is a fitting assignment for its engineer-turned-CEO. The code may change. The work is still discoverability.