The most revealing thing about Charu Sharma's new company is not that it uses artificial intelligence. In 2026, that claim hardly narrows the field. The revealing thing is where she has chosen to put it: inside the back office of a bank, among the investigations, internal policies, review queues and reports that exist because money dislikes mystery.
Fenrock AI, the company Sharma cofounded with machine-learning engineer Michael Mernagh, builds specialized agents for traditional financial institutions. The pitch is brisk: let software gather information, analyze cases and prepare work that analysts now assemble across disconnected systems. But the consequential word in the company's vocabulary is not agent. It is audit. The system is designed to leave a trail a human reviewer can follow.
That distinction suits Sharma. Her career has been an extended education in what institutions permit, whom they promote and how they decide a new tool deserves trust. She has built workplace mentorship software, helped scale infrastructure in a regulated industry, sat on the venture side of the table, and returned to the founder's chair. Fenrock, accepted into Y Combinator's Winter 2026 batch, looks like a fresh start. It also looks like the latest answer to a question she has been asking for years: how do you give capable people more leverage inside systems that cannot afford casual mistakes?
Chapter oneThe instinct before the résumé
Long before enterprise software, Sharma's name entered the public record for an act of instinct. In 2002, on a Mumbai local train, she and her younger brother Chinmay resisted armed robbers in a women's compartment. She held on to a pole when one attacker tried to force her from the train; her brother bit the attacker's hand. The children received India's National Bravery Award the following year. Their story later appeared in a Maharashtra school textbook.
Newspaper profiles naturally reach for adjectives such as fearless. Sharma's later account is more precise. “As a little girl, courage felt simple - it was instinctive,” she wrote two decades later. Adulthood, she said, taught her that courage also lives in mundane choices: speaking when something feels wrong, supporting an overlooked colleague, disagreeing with a decision that violates one's values. The dramatic episode is cinematic. Her definition of courage is operational.
The direction took her from India to Mount Holyoke College on scholarship. She studied physics and economics, a pairing that now reads like an early rehearsal: systems on one side, incentives on the other. Internships at young companies gave her real work while some classmates were still carrying coffee. At one of them, she encountered founders who shared her South Asian background. “If they could do it, then I could do it too,” she later recalled.
Representation, in this telling, was not a slogan. It was permission. The fact that those models were men soon sharpened the idea. Sharma created Go Against the Flow, first as a nonprofit project and book highlighting women founders, then as a documentary. She would go on to publish three books and speak widely, treating public storytelling as another way to make a path visible.
Chapter twoA mentor, multiplied
After college, Sharma joined LinkedIn. There she met a mentor, also an immigrant, whom she credits with changing her career. She tried to pass that benefit along by organizing a mentorship program for women inside the company. More than one hundred prospective mentors and mentees reportedly wrote to her within days. The surprise was the supply: there were people with experience who wanted to help, but the organization lacked a good mechanism for finding and matching them.
That observation became NextPlay.ai in 2017. Its mobile product, named Ellen, matched employees with mentors inside their companies and guided the relationship toward goals. Twitter, Intuit, Coca-Cola, Splunk and Logitech were among the organizations reported to have used it. The proposition was almost modest: mentorship already works; software can make access less dependent on luck.
NextPlay raised a seed round and won enterprise customers. It also closed in 2019. Sharma announced the end with an article carrying three blunt words: I closed my startup. “It's important to talk about failure!” she began. Silicon Valley usually dresses a shutdown in euphemism, preferably one that sounds like a strategic weather pattern. Sharma used a verb.
The closure matters because it prevents the tidy version of her career from becoming too tidy. Founder experience is not a ladder. It is a collection of scar tissue, customer calls and pattern recognition that can be carried into a different room. Years later, visiting Brex's headquarters, Sharma remembered spending long hours there during NextPlay's early customer discovery and demos. Startup time had folded back on itself. The company was gone; the learning was still in the building.
Chapter threeThe regulated apprenticeship
Sharma's next operating chapter was Osana Salud, where she served as cofounder and president. The company built back-end infrastructure used by institutions across Latin America, Spain and the United States. By the end of 2021, its network reached a reported six million people, and Forbes reported $26.5 million in capital raised. The scale earned Sharma a place on the 2022 Forbes 30 Under 30 list.
For the Fenrock story, the important lesson is not the list. It is the environment. Infrastructure buyers ask inconvenient questions. They care about reliability, integration, privacy and what happens on the dreary Tuesday when a system behaves strangely. Working in a regulated sector teaches a founder that the dazzling feature is only one guest at the procurement meeting. Controls, implementation and accountability have chairs too.
Sharma then stepped away from operating and joined SignalFire as an entrepreneur-in-residence and investor in 2022. She described the move as a chance to make a third startup more intentional. Venture capital offered distance from the daily emergency of one company and exposure to the decisions of many. It also made market timing a conscious part of her founder vocabulary.
By the time she returned with Fenrock, she had seen institutional software from most available angles: employee, customer discoverer, founder, executive, board participant and investor. The result is not caution exactly. It is a preference for ambitious technology wrapped in procedural respect.
The trust loop
The basic design choice: automate the assembly, preserve the responsibility.
The current betThe least glamorous room in the bank
Banks are attractive AI customers for the same reason they are difficult ones. Their back offices contain repeated research, document-heavy investigations and reporting work spread across legacy systems. The labor is expensive; the consequences of getting it wrong are more expensive. Fenrock overlays existing systems rather than demanding a wholesale migration, incorporates internal procedures, and prepares outputs for people to inspect.
Sharma's own formulation is useful: “Analysts should be verifying AI's output.” The sentence assigns the machine plenty of work without giving it the final word. It also reveals what Fenrock is really selling. Faster processing is easy to admire on a slide. A buyer in a regulated institution must be able to explain what the system did after the meeting is over.
The company says its agents can help investigate more fraud and money-laundering cases per analyst, speed lending work and resolve customer complaints. Those are different workflows, but they share a structure: information scattered across systems, policy that must be applied, a decision that needs evidence, and a human accountable for the result. Fenrock's product ambition is to become the workspace across that structure rather than another isolated tool.
Mernagh's background complements the wager. He spent a decade building machine learning at Google and Apple, including privacy-preserving systems. Sharma brings enterprise sales, regulated operations and the memory of what happens when a product's promise outruns its durable use. Their pairing makes the company less a generic AI story than a specific argument about where trust comes from: technical restraint, operational fluency and an interface that lets experts remain experts.
The through-lineAccess, followed by accountability
There is a pleasing symmetry between NextPlay and Fenrock. The first used software to connect employees with human mentors. The latest uses software to extend the reach of human analysts. Neither premise requires people to disappear. Both begin with expertise trapped behind a bottleneck and ask whether a product can distribute it more intelligently.
Sharma's public writing reinforces the pattern. She writes about failure without pretending it was secretly a triumph. She asks her network for questions before important meetings. She describes market timing as something to observe, not a mystical gift. Former colleagues emphasize her attention to their ambitions as well as company goals. The personality that emerges is social but not merely sociable: she treats networks as working systems through which people exchange confidence, information and opportunity.
Her childhood story will always sit at the front of the biography because drama has excellent manners and insists on being introduced first. Yet the quieter version of courage she later described may explain more. Company building consists mostly of ordinary decisions made under incomplete information: say the uncomfortable thing, close what is not working, learn a different market, return to the blank page.
Fenrock is still young. Y Combinator lists it as founded in 2026, active in San Francisco, with Sharma as founder and CEO. Its future will be decided by difficult, unphotogenic questions: whether agents can perform consistently, whether analysts trust the evidence, whether integrations survive real institutions, whether a buyer can defend the choice to a regulator. This is not a detour from the product. It is the product.
At seventeen, Sharma said she wanted to become an entrepreneur. She has now had enough versions of that job to know it is not one identity. It is a sequence of rooms: a dorm, a large technology company, a startup closing, a board table, a venture fund, a bank back office. Her career's distinctive move is to enter each room asking who lacks access and what the institution needs before it will say yes. Fenrock's answer is software that moves quickly, shows its work and leaves a person in charge.