After spending eight years developing advanced artificial intelligence systems for Meta, Rohit Patel is stepping away from the tech giant to chart a new entrepreneurial course. Patel, who formerly served as a director at Meta Superintelligence Labs, has launched Ndoc, an early-stage startup designed to function as a complete, AI-native alternative to legacy workplace mainstays like the Microsoft Office suite and Google Workspace.
Patel’s departure places him among a notable wave of high-profile exits from Meta’s upper ranks. His departure follows the recent departures of Chief Global Affairs Officer Joel Kaplan, prominent AI researcher Andrew Tulloch, and Emily Dalton Smith, who served as Meta’s head of product for AI-for-work transformation. While tech giants continue to shuffle their leadership and engineering teams in the hyper-competitive artificial intelligence race, Patel is betting that the future of enterprise software does not lie in patching AI onto decades-old desktop metaphors, but in reimagining the digital workplace from the ground up.
Unlike many contemporary productivity tools that simply overlay an AI chat interface or assistant onto an existing spreadsheet, word processor, or presentation software, Ndoc was designed from its inception to eliminate the friction of fragmented workflows. Modern knowledge workers frequently find themselves bouncing between an exhausting array of distinct applications—shifting data manually across Excel, Word, PowerPoint, Tableau, Replit, and relying on endless cycles of copy-pasting from standalone LLM chats like ChatGPT. Ndoc seeks to consolidate this scattered ecosystem into a single, unified workspace that can be accessed and utilized by colleagues, even if those team members are not active subscribers to the platform.
The architectural layout of Ndoc reflects a deliberate effort to synthesize different modalities of human work into one cohesive interface. The platform divides its operational layout into three core components. The first is the Stage tab, which serves as the home for presentations of key findings, encompassing slide decks, interactive dashboards, simulations, and lightweight mini-applications. The second is the Narrative tab, dedicated to deep-dive analyses, numbered citations, dynamically generated figures, and the contextual reasoning behind raw data—the underlying "why" that explains the numbers. Finally, the Model tab houses the foundational datasets and Python scripts that drive the computations.
This structural design directly mirrors Patel’s professional background and research focus during his tenure at Meta. While at the social media giant, Patel concentrated heavily on reinforcement learning—a branch of machine learning where AI agents learn to optimize decision-making processes through trial and error—as well as the critical challenge of evaluating whether artificial intelligence outputs are genuinely trustworthy. Patel firmly believes that if enterprises are going to trust autonomous systems to draft the spreadsheets and slide decks that drive major corporate decisions, human supervisors must be able to audit and verify the system’s work. By exposing the underlying Python scripts within Ndoc, users can trace the exact logic used to generate a report.
Even for professionals who do not write code, Ndoc aims to make verification accessible through conversational interaction. Users can ask the platform’s internal chat interface to clarify a script, locate a specific formula line, and confirm its mathematical integrity. Patel emphasizes that the system operates with absolute transparency, noting that nothing is hidden behind a proprietary black-box curtain.
In everyday practice, the workflow is streamlined: a user uploads a raw dataset, prompts Ndoc to execute an analysis, and instructs the system to draft a comprehensive narrative explaining the findings. Furthermore, the platform allows users to inject document-specific context regarding the target audience, preferred tone, and long-term memory. According to Patel, the system preserves context and memory continuously across individuals, specific documents, entire project teams, and the broader enterprise.
Under the hood, Ndoc is constructed and fine-tuned using a careful mixture of open-source models. Because the platform was built entirely from scratch rather than being chained to massively expensive proprietary foundation models, Patel notes that it costs considerably less to operate than competing solutions. Interestingly, while Patel initially anticipated that this cost-efficiency would be the primary selling point for prospective clients, early user feedback revealed a different priority. When testing the software with initial users, Patel found that enterprise customers cared far less about the underlying model economics and far more about the sheer convenience of having their entire end-to-end analytical workflow housed in a single location.
Patel’s broader ambition extends far beyond building a better desktop application; he aims to fundamentally reinvent the digital file format itself. Ndoc allows documents to be exported as a self-contained .ndoc.html file that can be opened directly within any standard web browser. Patel plans to open source this proprietary file format so that independent developers can build third-party applications that integrate seamlessly with it, breaking the historical cycle of vendor lock-in that has tied enterprise customers exclusively to Microsoft or Google ecosystems for decades.
This desire to champion open-source standards was a primary catalyst behind Patel’s decision to leave Meta and launch his own venture. Operating inside a massive corporation often requires navigating labyrinthine layers of bureaucracy, where achieving consensus on a universal open-source format would necessitate alignment from hundreds of stakeholders holding divergent opinions. Eager to bypass organizational friction, Patel chose independence to maximize his velocity. He notes that humans are inevitably bound by human operational speed, whereas building alongside artificial intelligence allows him to operate at machine pace.
Attempting to capture enterprise workflow market share is a high-stakes gamble for a lean, bootstrap startup operating with a tiny team and no external venture capital funding, especially given that major AI labs are aggressively targeting the exact same territory. Industry heavyweights including OpenAI, Anthropic, Meta, SpaceXAI, and Google have all rolled out enterprise-focused capabilities designed to tether their proprietary models directly into Google Workspace, Microsoft 365, and other legacy workplace infrastructure. Notably, OpenAI recently launched Space, a shared workspace integrated directly inside ChatGPT that represents an aggressive, direct challenge to established office productivity suites.
While Patel acknowledges that products like OpenAI’s Space present a formidable competitive threat, he contends that established AI labs are fundamentally misinterpreting how professionals actually need to work. In his view, existing enterprise AI efforts are superficial, merely attaching generative features to traditional spreadsheets to automate tedious cell-filling tasks. He compares this approach to discovering electricity and concluding that the best possible application is strapping lightbulbs to oxen plowing a field, arguing instead that the industry needs a proper tractor.
At the same time, Patel recognizes the immense gravitational pull of established habits, which explains why major AI labs have prioritized bringing their models directly into pre-existing workflows. To bridge this gap without forcing immediate, painful behavioral changes, Ndoc includes robust connectors that allow customers to ingest and import existing data from the software tools they already rely on daily. Patel maintains a strict personal benchmark for his software, stating that he refuses to release any feature or workflow that he does not personally love using. Having accumulated a professional graveyard of tools that failed to meet his rigorous daily productivity standards, he insisted on rigorous internal dogfooding before bringing Ndoc to the public.
That uncompromising standard extends to customer acquisition and funding strategy. Patel deliberately chose to build the product to a functional state before seeking financial backing, preferring tangible proof of market demand over speculative pitch decks.
As of Tuesday, Ndoc has officially initiated an enterprise pilot program while opening an alpha version to a limited group of users via a waitlist. While signing up for the platform is free, customers will operate on a usage-based pricing model tied to token consumption. Patel anticipates rolling out a broader beta version in the coming weeks, followed by a full commercial launch next year. For Patel, maintaining relentless momentum and prioritizing rapid execution outweigh the pursuit of an immaculate initial product release, operating on the philosophy that by the time a company perfects its grand plan, someone else will have already executed it.