ProfileFour companies, one architectural instinctPipeLinks → Aarohi → Lightwire → Baffle100B+ records protectedAssume the breach. Protect the data.Four companies, one architectural instinct

Founders / Data Security / The Long Game

Ameesh Divatia's Long Game: Make the Data Useless to Thieves

After three acquired companies and a spell as an investor, Ameesh Divatia returned to the operator's chair with a stubborn premise: a breach matters less when the stolen data reveals nothing.

The small object on the magazine cover was a microchip. Ameesh Divatia was a boy in India, looking at an image in National Geographic and trying to imagine what such a tiny thing could eventually do. Another cover followed: TIME's 1982 choice of the personal computer as its “Machine of the Year.” The machinery was interesting. The larger possibility was irresistible.

Decades later, the same mental move still organizes Divatia's work. Start with a component. Pull the camera back. Ask what system it might change. He has carried that habit through networking silicon, optical networking, storage, silicon photonics and cloud security. The markets kept moving. His favorite question did not.

“I always want to know what the big picture is.”Ameesh Divatia

Divatia describes himself as a “decent student,” then supplies his own calibration: somewhere in the top five percent. He played sports, watched his father move from pharmacology into corporate leadership, and studied electronics at M.S. University of Baroda. At 22, directly out of college and without a day of full-time work experience, he moved to California for graduate school.

At UC Irvine he concentrated on the performance analysis of communication networks. An internship supplied a useful shock: school and work were different species. The equations mattered, but customers and systems decided what survived outside the classroom. That distinction would later become one of his recurring rules for founders.

01 / The portable idea

Cross the border next door

His early jobs at Advanced Micro Devices, Allied Telesyn and 3Com moved him from application engineering toward architecture. He wanted the broad view of how a customer used a product. Seven years after graduate school, during the bright, excitable summer of 1996, he quit to start PipeLinks.

The idea took knowledge from data networking and carried it sideways into optical networks. PipeLinks built routers that could transport traditional circuit-based voice and internet traffic at the same time. Cisco, already an investor, agreed to acquire the company in December 1998 for approximately $126 million in stock. Divatia joined Cisco as chief product strategist for its optical transport business.

He has a name for this move between neighboring fields: cross-leverage. Find a lesson that has become normal in one market. Apply it to another market that has not changed in a while. The phrase sounds like consulting language until you notice that it became a repeatable company-building method.

Ameesh Divatia's cross-leverage map A sequence from data networking through optical networking, storage networking, silicon photonics and data protection. DATANETWORKS OPTICALNETWORKS STORAGENETWORKS SILICONPHOTONICS DATAPRIVACY
The Divatia method: borrow a mature idea from one technical neighborhood and make it useful in the next.

In 2001, storage was the next border. Divatia founded Aarohi Communications, assembled a team and raised seed capital, then led the company into intelligent processors and software for data-center networks. Emulex bought Aarohi in 2006 in a deal valued at up to roughly $39 million. The acquirer was not merely purchasing a product. It was opening routes into 10-gigabit data-center networking and intelligent storage platforms.

02 / The partner's math

One dollar here, ten dollars there

Founders can become besotted with their technology. Divatia's correction is unsentimental: “Nobody cares about technology per se; it has to solve a problem.” His companion idea is a multiplier. If his business is worth one dollar to him, it should be worth ten dollars to the company on the other side of a partnership. The founder has to understand where that extra value comes from.

That logic was visible again at Lightwire. The silicon-photonics company had technically ambitious work and heavy capital needs. Divatia has said the team looked for a “big brother” that could help productize it. Cisco invested, supplied a demanding specification for throughput on a router blade, and worked with the company. Lightwire's team met that target in less than two years. Cisco acquired it in 2012 for $271 million.

3Companies led through acquisition before Baffle
$436MCombined announced transaction values
100B+Records Baffle says its software protects

The common thread was not an obsession with selling companies. It was an insistence that technical value become strategic value for a larger system. Divatia advises founders to build relationships with prospective acquirers over time, to share the value proposition without exposing the crown jewels. An exit, in this telling, is a long conversation before it is a transaction.

The sequence looks clean in retrospect. Building never is. Divatia is quick to invoke luck and teams. He says he has always worked with people better than himself. The humility is paired with a fairly demanding standard - for the idea, for the team and for his own grasp of the customer's problem.

03 / Back into the arena

The investor who missed operating

After Lightwire, Divatia wanted a break from operating. With his former CFO and a mentor, he started the seed fund Carta Ventures. It invested in eight companies; by 2024, he said six had been acquired. The results did not resolve a more personal problem. He grew bored watching entrepreneurs run their companies.

Then an early-stage investor introduced him to Priyadarshan “PD” Kolte and posed a question: how can people get value from data while still protecting it? The contradiction had the right shape. Divatia likes difficult technical problems, especially ones whose practical moment has arrived. He and Kolte co-founded Baffle in 2015.

The name contains the architecture. Imagine software processing sensitive information while the administrator of the underlying infrastructure cannot see what the information says. The administrator is baffled. More usefully, an intruder who steals the administrator's credentials may also be baffled.

“You have to be open about sharing your value proposition, but not your crown jewels.”A founder's rule for strategic relationships

Traditional enterprise encryption has protected data while stored and while moving between systems. The difficult interval is use, when a database or application has to work with the information. Baffle places a software layer between applications and data stores, applying masking, tokenization or encryption at the field level. Its proxy approach is designed to avoid rewriting the application.

The Baffle data-protection flow An application sends a query through a protection layer to encrypted data, while different users receive policy-based views. APPLICATIONNO REWRITE PROTECTIONLAYERMASK · TOKENIZE · ENCRYPT ENCRYPTEDDATA POLICY DECIDES WHO SEES WHICH VIEW
Compatibility is part of the product: Baffle's layer is meant to keep existing applications working while policy changes what each user can see.

The practical beneficiary is not always the person who finds the product. Divatia says security teams usually discover Baffle, while data scientists benefit from being able to work under rules that protect sensitive values. It is a revealing product insight. The buyer, operator and beneficiary can occupy three different desks.

Steal this operating principle

Map the whole system around a product: who discovers it, who approves it, who runs it, who benefits, and whose workflow must remain unchanged.

Baffle says its technology now protects more than 100 billion records, integrates with AWS, Azure, IBM and Google Cloud, and has raised more than $35 million. Its 2021 Series B brought in $20 million. In 2023, the company described a yearlong collaboration that put its protection capabilities at the center of an IBM data-security broker. More recently, it extended its work into Amazon S3 ingestion and protection for AI data pipelines.

04 / The life outside the diagram

Good, bad and ugly

The systems thinker does not pretend organizations are only systems. Divatia describes his leadership style as open: tell the team the good, the bad and the ugly. He refuses sole credit and emphasizes teamwork. At home, he tries to leave work at the office. Compartmentalization is his preferred boundary.

Sports supply another compressed model of the world. He plays golf and cricket and follows professional competition broadly. In a game, he has observed, the full emotional range of life arrives in two or three hours. Travel stretches the perspective in the opposite direction. He has visited every continent except Antarctica and likes the history embedded in old places - evidence of what people built before they had modern tools.

That curiosity helps explain the fourth startup after three acquisitions and an investor's intermission. Divatia has said startups let him go home knowing he accomplished something without depending on someone else to make it happen. The remark is less swagger than preference. He likes being responsible for moving the piece.

PipeLinks

Optical networking, acquired by Cisco.

Aarohi Communications

Storage networking, acquired by Emulex.

Lightwire

Silicon photonics, acquired by Cisco.

Carta Ventures

A seed investor's intermission.

Baffle

Data protection designed around use, not just storage.

Security vendors often sell fear. Divatia prefers the positive case: disciplined data handling can become a competitive advantage. His deeper bet is architectural. Cloud systems have made data more useful by making it easier to copy, combine and analyze. The same fluidity creates exposure. Instead of hoping every wall holds, protect the value moving through the rooms.

The premise is stubborn because it accepts an unpleasant fact. Credentials will be compromised. Administrators may be impersonated. A perimeter will fail. Start there, then reduce what failure can reveal. For someone who began by asking how a microchip could alter the whole system, it is a familiar move: focus on the smallest field of data, then pull the camera back until the consequence becomes clear.