A job description went up. One hundred and twenty applications came back. The founder had neither the hours nor the appetite to read every résumé, run fifty screens and slowly compress the pile into a shortlist. So Abhishek Agarwalla connected the role to Fabric, the recruiting product he was building. The software screened the résumés, sent interview links, conducted two dozen conversations and returned six people worth his attention. Agarwalla spoke with three and liked two. His target was a seven-day trip from job description to offer, with six or seven hours of his own time.
He described the episode with a wink: using Fabric to hire for Fabric felt like getting high on his own supply. Beneath the joke was a clean product test. A founder had put his own team, calendar and judgment behind the output. The software would not be graded on how smoothly it moved through a demonstration. It had to find people he might actually hire.
That loop captures Agarwalla's career. He learns a system by working inside it, notices where the signal degrades, then builds around the bottleneck. Petroleum reservoirs gave way to machine-learning models. Client projects became an AI consultancy. Years of hiring for that consultancy became a recruiting company. Each turn looks abrupt from a distance. Up close, it is the same habit repeated.
A petroleum degree, a systems education
Agarwalla arrived at the Indian Institute of Technology (Indian School of Mines), Dhanbad in 2013 to study petroleum engineering. He spent four years learning about pressure, extraction and complex systems that fail in expensive ways. His research work included CO2-based fracturing, proppant optimization and tight gas reservoirs. He presented it at international exploration and production conferences.
College also pulled him toward programming, writing, music and startups. During a 2014 AIESEC exchange in Mauritius, he volunteered on environmental campaigns, worked as a blogger and photographer, and helped with coral reef conservation and pollution awareness. The combination is revealing: technical study on one side, communication and field work on the other.
When he graduated in 2017, he did what he later called the perfectly rational thing for a petroleum engineer: he went into AI. At ZS in Pune, he worked as a data scientist on predictive models and commercial analytics. The job put him near applied machine learning, where a model has to survive contact with a real business decision.
The consultancy was the classroom
In July 2018, Agarwalla co-founded Aidetic with college friends Mehul Jain, Ketan Mishra and Yadvendra Kshatri. The group had been friends, roommates and colleagues before becoming business partners. The first work centered on computer vision and video analytics. Early projects included a reconnaissance system for a drone company and security systems for schools. The company moved from Pune to Bengaluru and widened its range.
When the pandemic made physical deployments difficult, Aidetic moved deeper into natural-language processing and transformer models. It worked across retail, financial services, media, sports and enterprise software. One project involved a conversational voice assistant. Another concerned recommendation systems for online retail. The team also applied its skills to sports analytics, Braille recognition, quality-control systems and data pipelines.
By 2022, Aidetic had reached $1 million in revenue with more than twenty active clients. The team passed sixty people the following year. Over its run, the company delivered more than 300 AI projects across more than fifty clients. Scale changed Agarwalla's job. He had started close to execution, then acquired responsibility for the tactical layer and eventually the strategic one.
He resisted becoming a leader who worked only in abstractions. In a long interview about Aidetic, he described management as three connected levels: execution at the bottom, tactical coordination in the middle, and strategy at the top. He still participated in sales closing because it kept him in contact with buyers and implementation. His operating ideal was to move between layers according to where the company needed attention.
“Your customers remember how you made them feel far, far longer than your feature list.”Abhishek Agarwalla
That instinct became visible in a small Fabric story. A customer emailed about an irritating tooltip bug at 12:03 p.m. The fix was in production at 12:59. Agarwalla presented the fifty-six-minute response as evidence of the company's promise to care about every customer. It is an unglamorous anecdote, which is why it works. Culture lives in the queue: what gets noticed, how quickly someone owns it, and whether the user has to repeat the problem.
12:03 - customer reports a tooltip bug. 12:59 - the fix is deployed. A company value measured with a clock.
Hiring becomes the product
Building Aidetic meant hiring engineers, salespeople and managers. Agarwalla estimates that he and his collaborators accumulated more than 10,000 hours doing it. The repeated work produced a frustration: résumés offered noisy summaries, agencies consumed time, and experienced employees were repeatedly pulled away from their work to conduct first-round calls.
The seed of Fabric appeared before the company did. Aidetic had built a voice assistant that could hold natural-sounding conversations, initially to help process job applications and potentially to support customers. By January 2025, Agarwalla and Ketan Mishra had co-founded Fabric in San Francisco around the recruiting problem itself.
Fabric began with an AI interviewer and expanded into what the company calls an AI Hiring OS. A recruiter can describe a role, source candidates, run outreach across email, phone and WhatsApp, screen applications, schedule calls and conduct first-round interviews. Transcripts and scorecards return to the existing applicant tracking system. The hiring team remains responsible for the final decision.
The distinction is central to Agarwalla's public argument. He builds AI for recruiting while insisting that recruiters are not disappearing. His claim is about time allocation. If software handles résumé review, scheduling and repeatable screening, recruiters can spend more time assessing context, building relationships and helping candidates after an offer.
There is also a claim about signal. Agarwalla argues that AI-assisted résumé writing makes applications increasingly difficult to evaluate. Fabric's response is a conversation that can probe skills, ask follow-ups and preserve a transcript. It is a bet that structured interaction can reveal more than polished prose. The product also checks behavioral signals for possible cheating, an issue that led the team to create OpenRound in 2026.
“Recruiters are not going away anywhere. AI or not.”Abhishek Agarwalla
A work trial after the interview
OpenRound extends the same search for evidence. Candidates receive an ambiguous problem, a real codebase and access to coding tools, then are evaluated on what and how they ship. It looks less like a quiz and more like a small piece of the job. Agarwalla co-founded the product in March 2026, launched it publicly in April and joined the spring Canopy cohort at Founders, Inc. in San Francisco.
The sequence from Fabric to OpenRound is a response to a moving target. As candidates gain access to better AI tools, assessments designed to exclude those tools become brittle. A work trial can instead ask whether someone uses them well. Agarwalla's concern remains consistent: replace weak proxies with observable behavior.
Fabric, meanwhile, announced a $110,000 angel round in October 2025. At the time it named Meesho, CRED, Kearney and MakeMyTrip among its enterprise customers and said the platform had processed thousands of interviews. Published customer results later described one engineering campaign that screened more than 10,000 candidates and saved more than 1,000 senior engineering hours.
Numbers at that scale make the argument for automation easy. The harder argument concerns trust. Candidates need to know what is being assessed. Recruiters need evidence they can inspect. Hiring managers need control over the decision. The product has to reduce work without hiding the basis for its recommendation.
Close enough to feel the friction
Agarwalla's public style is blunt, numerical and occasionally mischievous. He writes about growth rates, applicant counts, sales leads and time saved, then punctures the spreadsheet with a joke about his wedding or his own product habit. The humor makes the operating pressure legible. Fabric was growing, he wrote in one hiring post, while fifteen-hour days and a crowded sales calendar competed with plans to get married.
His collaborators recur across companies. Jain, Mishra and Kshatri were friends before Aidetic. Mishra later co-founded Fabric. Longstanding college ties also helped open the conversation that led to Aays acquiring Aidetic. The network is less a collection of contacts than a group of people who have repeatedly built together.
The through line is proximity. At ZS, Agarwalla worked on models connected to commercial decisions. At Aidetic, he kept closing sales while learning strategy. At Fabric, he used the product to hire his own employees and treated a tiny customer bug as an immediate operating problem. OpenRound brings evaluation even closer to the job by watching a candidate work.
There is no clean endpoint to hiring. Tools change, candidates adapt, and yesterday's signal becomes tomorrow's script. Agarwalla appears comfortable with that instability. His career has been a series of systems entered from the ground floor. The useful lesson is portable: stay close enough to the work that recurring annoyance becomes visible. Then build from evidence, and keep the consequential human judgment in view.