After dedicating eight years to developing artificial intelligence at Meta, Rohit Patel has officially departed the tech giant to launch his own ambitious venture. TechCrunch has exclusively learned that the former director at Meta Superintelligence Labs has founded Ndoc, a startup designed from the ground up to serve as an AI-native replacement for traditional productivity suites like Microsoft Office and Google Workspace.
Patel’s exit is part of a broader wave of high-profile departures from Meta’s upper ranks. His departure follows the recent exits of several key leaders, including Chief Global Affairs Officer Joel Kaplan, prominent AI researcher Andrew Tulloch, and Emily Dalton Smith, the former Head of Product for AI-for-Work transformation. These transitions highlight a shifting landscape within major technology companies as top talent branches out to pursue independent visions in the rapidly evolving generative AI sector.
Rethinking the Enterprise Workflow
Rather than simply layering AI capabilities onto legacy tools or existing document formats, Patel designed Ndoc to fundamentally rethink how professionals handle complex workflows. For years, office work has been fragmented across a sprawling collection of disparate applications, forcing employees to constantly toggle between spreadsheets, word processors, presentation software, data visualization tools, coding environments, and traditional chatbot interfaces where data is manually copied and pasted.
Ndoc aims to eliminate this friction by consolidating the entire lifecycle of enterprise work into a single, unified workspace. Crucially, the platform allows colleagues to access and interact with these workspaces even if they do not maintain an active Ndoc account themselves, lowering the barrier to adoption across cross-functional teams.
The layout of Ndoc reflects Patel’s technical background and his specific philosophy regarding artificial intelligence in the workplace. During his tenure at Meta, Patel focused extensively on reinforcement learning—a specialized machine learning methodology where AI agents learn to make sequential decisions through trial and error to maximize rewards—as well as the critical challenge of evaluating whether AI-generated outputs are genuinely trustworthy.
Patel firmly believes that if artificial intelligence is going to construct the financial models, spreadsheets, and slide decks that businesses rely on to make high-stakes decisions, human users must be able to easily audit and verify the work. To achieve this transparency, Ndoc’s interface is divided into three distinct, interconnected components housed within a single layout.
A Unified Three-Tab Layout
The first component is the Stage tab, which serves as the visual home for presentations of key findings, including slide decks, interactive dashboards, simulations, and mini applications. The second component is the Narrative tab, which houses the written analyses, numbered citations, generated figures, and the contextual "why" behind the numbers. The third component is the Model tab, where the underlying datasets and Python scripts powering the analysis are stored and maintained.
This architecture ensures that users are never left wondering how a specific conclusion or data point was reached. Even individuals who lack programming experience can interact directly with the platform’s AI chat interface to inspect the code. Users can ask the AI to explain specific scripts, locate particular formulas, and verify calculations line by line. According to Patel, this design philosophy ensures there is nothing hidden behind the curtain, granting professionals total visibility and control over their automated outputs.
In practice, the platform allows users to upload raw datasets and prompt Ndoc to perform comprehensive analyses while drafting a narrative report detailing the findings. Furthermore, users can enrich these documents with specific contextual parameters, such as the intended audience, preferred tone, and persistent memory. Patel notes that the system maintains a continuity of memory across individual users, specific documents, entire teams, and the broader enterprise.
Open-Source Foundations and Reinventing the File
Under the hood, Ndoc is built and fine-tuned using a mixture of open-source models rather than relying exclusively on costly proprietary frontier models. Because the platform was engineered entirely from scratch, Patel points out that it costs considerably less to operate than competing alternatives. Initially, he anticipated that this cost efficiency would be the primary selling point for prospective clients. However, early trial users have demonstrated that their primary enthusiasm lies elsewhere—specifically in the convenience of viewing their end-to-end workflows in a single, cohesive location.
Patel’s long-term ambition extends beyond the workspace interface itself; he ultimately aims to reinvent the fundamental concept of the digital file. Ndoc allows documents to be exported as a single .ndoc.html file that can be opened and interacted with directly within any standard web browser. Patel plans to open-source this proprietary file format, enabling other developers to build third-party applications that integrate with it seamlessly. This open ecosystem approach stands in stark contrast to the historical model enforced by tech giants like Microsoft and Google, which have traditionally locked users into proprietary ecosystems for decades.
This deep commitment to open-source development was a primary catalyst for Patel’s decision to leave Meta and build his own company independently. Navigating an open-source initiative of this magnitude within a massive corporate structure like Meta would inevitably require navigating endless rounds of review, involving hundreds of stakeholders and competing opinions on structural design. Opting to go it alone allows Patel to operate at the speed of technological capability rather than traditional human consensus.
Navigating a Crowded Enterprise AI Market
Attempting to establish a dominant enterprise platform is a monumental undertaking for a bootstrapping founder operating with a lean team of three and zero outside funding, particularly at a time when major AI labs are fiercely contesting the exact same territory. Industry heavyweights including OpenAI, Anthropic, Meta, SpaceXAI, and Google have all rolled out enterprise-grade AI capabilities designed to connect directly with established productivity suites like Microsoft 365 and Google Workspace. Furthermore, OpenAI recently launched Space, a shared workspace integrated directly within ChatGPT that poses a direct challenge to traditional office applications.
While Patel acknowledges that well-funded competitors like Space represent a significant competitive threat, he contends that major AI labs are approaching the problem from the wrong angle. In his view, existing enterprise AI tools merely attach automated assistance to legacy interfaces, automating minor tasks like filling out spreadsheet cells. Comparing this strategy to misplaced technological adaptation, Patel remarks that using advanced AI merely to populate spreadsheet cells is akin to inventing electricity and deciding the best application is strapping electric lights to oxen plowing a field, rather than building a tractor.
At the same time, Patel recognizes the immense gravitational pull of established habits, acknowledging why major AI labs have prioritized bringing their models directly into the tools where workers already spend their time. To bridge this gap, Ndoc features robust connectors that allow customers to effortlessly import existing data from their legacy productivity tools. Patel maintains a strict personal standard for product viability, stating that he refuses to release anything he would not eagerly use himself. Having accumulated a personal graveyard of tools that failed to meet his rigorous bar, he ensured Ndoc passed his own heavily optimized daily workflows before introducing it to the market.
Moving Fast and Executing Without Funding
True to his philosophy of rapid iteration, Patel adopted a strict benchmark for customer validation before seeking venture capital funding. He deliberately chose to avoid raising external capital until he had tangible proof that real users genuinely loved the product.
As of Tuesday, Ndoc has officially commenced an enterprise pilot program while opening up an alpha version to a restricted group of customers drawn from its waitlist. While signing up for the platform is free, customers operate on a usage-based token model. Patel anticipates releasing a broader beta version in the coming weeks, followed by a full commercial launch planned for next year.
For Patel, maintaining relentless momentum and prioritizing swift execution far outweighs the pursuit of initial perfection. In his view, spending excessive time crafting a flawless long-term plan is a liability when execution can be achieved immediately, noting that by the time a perfect plan is finalized, the work will have already been completed by someone else.
