Abhi Sharma Maps the Things We Cannot See

The first map Abhi Sharma built was meant to help people find their way around a building. He was a graduate student at Carnegie Mellon, caught between the orderly future of a doctorate and the messier pull of industry. A professor offered a practical test: build something and see where his heart went. Sharma turned an independent study into a small company called Kontext, recruited another engineer through the same academic loophole, and used Bluetooth beacons to trigger indoor directions and location-aware offers. People used the product at Lowe's. The experiment answered his question. He liked making systems that escaped the laptop and became useful to strangers.

Years later, Sharma would build another kind of map. This one would show where customer data begins, which services touch it, how it crosses a cloud environment, where an outside vendor receives it, and whether any of that motion agrees with a contract or policy. The scale changed. The instinct did not. A hidden system becomes negotiable once everyone can see it.

The path between those maps passed through AppDynamics, where Sharma worked as a platform engineer, and FogHorn Systems, an edge-AI company he helped build as a founding team member. He worked across compilers, real-time processing, machine learning, analytics, and observability. One patent from that period describes how pattern-driven reactions can be translated for real-time dataflow computing. The vocabulary is technical, but the recurring concern is plain: events happen quickly, in sequence, and software needs a way to understand the pattern.

Pizza, privacy, and the missing layer

Late in 2019, Sharma returned from climbing Mount Kilimanjaro and met his longtime friend Leila Golchehreh for pizza in San Francisco. His previous startup was moving toward an acquisition, and he was searching for the next problem. He had also been reading about progress studies and was drawn to questions that sit between established fields. Golchehreh, a privacy attorney and operator, had spent years building legal and data-protection programs. She knew the daily burden of the tools.

Lunch stretched to roughly eight hours, with the discussion migrating through several coffee shops. They talked about privacy, artificial intelligence, law, and the distance between a policy and the software it was meant to govern. In Sharma's retelling, the combination was almost comic: burrata on pizza, a privacy lawyer, and a compiler nerd. Yet the mismatch was the point. Golchehreh understood obligations from the top down. Sharma knew how to inspect a system from the bottom up.

Their shared observation was that privacy teams were working from periodic answers while engineering teams were changing the underlying reality every day. A questionnaire might record what an application did in March. By April, new code, services, models, and vendors could make the record stale. The common approach moved compliance work from spreadsheets into browser forms without closing that time gap.

Relyance AI, founded in 2020, was their attempt to build the missing layer. Its software connects to code repositories, infrastructure, cloud systems, applications, and business tools. It produces an inventory and a visual lineage of data, then relates what is happening to contracts, regulations, and internal commitments. Legal can see the operational system. Engineering can see why a flow carries an obligation. Security can see where movement creates exposure.

Sharma likes the image of a nervous system. In a large company, no single person knows every service or every path between them. A map gives each specialist a way into the whole. The company calls its later version of this idea Data Journeys: a record concerned with where data came from, what it touched, why it moved, and where it may go next.

When the bug became the demo

The most revealing Relyance story happened during its first deployment for a paying customer. The team had spent eight or nine months building what Sharma calls a beta-plus product. At eight in the morning, lawyers, engineers, and a security specialist joined the call. The customer's tenant was empty. They connected infrastructure monitoring and deployed a container for static code analysis.

Then the visualizer lagged.

Sharma wanted the graph to render in one clean shot. Instead, assets appeared one by one and lines began joining them, like a string of lights switching on. From his side of the call, the delay looked like an engineering flaw. From the customer's side, the system seemed to be discovering the organization in real time. An engineer spotted a flow they had not known existed.

In that instant, the product did exactly what its founders had promised. The lawyer, the security professional, and the engineer were looking at the same fact. The customer purchased the product and negotiated pricing for the following two years. Sharma has said he will never forget the deployment.

The useful product lesson lives in his reaction. He still talks about the magic of watching the map appear, even after years of demos. The graph is not decoration. It turns an argument about governance into a shared object. A team can point, ask why, and decide what to change.

Money followed the map. Relyance emerged from stealth in 2021 with a previously unannounced $5 million seed round and a $25 million Series A. Jyoti Bansal, the AppDynamics founder who had stayed in touch with Sharma, led the seed investment through Unusual Ventures and joined the board. Menlo Ventures co-led the Series A with Unusual. In 2023, Sharma pitched the company as an RSAC Innovation Sandbox finalist.

In October 2024, Relyance announced a $32.1 million Series B led by Thomvest Ventures, with Microsoft's M12, Cheyenne Ventures, Menlo, and Unusual participating. At the time, the company said customers included Coinbase, Snowflake, MyFitnessPal, and Plaid. The funding gave Sharma a larger version of the old founder problem: a product can map a complicated organization, but the founder also has to build one.

Making culture observable

Sharma calls himself a second-time founder, and his public discussions about Relyance's growth are candid about the cost of delayed clarity. He has described a period when hiring mistakes, strategy disconnects, and weak enablement surfaced after the company had reached several million dollars in annual recurring revenue. Culture had existed as intent. It had not been translated carefully enough into behavior.

His response borrowed a phrase from the restaurant world: unreasonable hospitality. At Relyance, the idea means making an experience personal enough to be memorable and useful. Sharma has discussed small examples, including sending a tailored video to help win a competitive deal and putting a humidifier in an interview room when a candidate needed one. Another hiring story involved flying to meet a candidate who had recently become a parent instead of asking the candidate to travel.

The gestures are modest by startup standards. Their value comes from specificity. A slogan cannot notice that someone needs humidity or that travel is unusually hard that week. A person can. Sharma built the principle into rituals, internal stories, feedback, and recognition. He has also said he leads monthly culture onboarding for new hires himself.

There is a symmetry between this management philosophy and the product. A regulation only becomes useful when connected to an actual data flow. A cultural value only becomes useful when connected to a choice someone can make on Tuesday afternoon. Sharma's preference is for operational evidence.

The map starts moving

Artificial intelligence has made Sharma's original timing problem more severe. Conventional software follows instructions written in code. AI systems also respond to the data and context around them. Agents can call tools, use service identities, create temporary infrastructure, reach third-party models, and move information through a chain that may exist for minutes.

Sharma's current argument is that security teams have become good at asking object questions. Where is the database? Which bucket contains personal information? Who has permission? Those facts still matter. The exposure often appears only in the interaction: an agent with broad access, an unvetted tool, a sensitive dataset, and a temporary destination connected in sequence.

In March 2026, Relyance made Lyo commercially available at RSAC. The company presents it as an autonomous data defense engineer that monitors data activity across code, cloud infrastructure, identities, software services, third parties, and AI agents. Sharma's line for the launch was concise: “Context is the only thing that separates a scanner from true defense.”

It is a commercial thesis, naturally, but also the mature form of his old obsession. An indoor location has limited meaning until a person wants to get somewhere. A data asset has limited meaning until a system touches it for a purpose. The more useful map includes time, intent, and movement.

Sharma's career also contains reminders that the builder behind this rather abstract machinery is attentive to feeling. He has said he could listen to Amy Winehouse's rendition of “Valerie” forever because it sounds as if it came from the heart. He chooses dogs over cats. He climbed Kilimanjaro, then returned and spent most of a day talking through privacy law over pizza. His extended work title even adds “Chief Leaf Blower” after founder and CEO. These are small details, but they keep the systems story grounded.

The aspiration now is larger than privacy automation. Sharma talks about trust and governance infrastructure for an era in which software has more agency and data can function like an instruction. The company has expanded from privacy operations into AI governance and data security. Its customers must decide whether that unified approach replaces the collection of specialized tools they already own.

For Sharma, the answer will be proved in the same place as the first one: on the screen, in the room, when a line appears between two things that no one had connected before. Good maps do more than describe a territory. They change the conversation among the people standing in it.