Meta AI director Rohit Patel leaves to found Ndoc, a challenge to Office and Workspace
Rohit Patel, former AI director at Meta Platforms, has left to found the startup Ndoc, positioning it as an AI-native alternative to the Microsoft Office suite or Google Workspace. TechCrunch reported exclusively on October 27, 2026, that Patel worked at Meta for eight years and most recently served as director of Meta Superintelligence Labs, mainly responsible for reinforcement learning and evaluating the trustworthiness of AI outputs.
Ndoc aims to integrate workflows currently scattered across Excel, Word, PowerPoint, Tableau, Replit, and the “ChatGPT copy-paste” workflow into a single workspace. The product includes three components: Stage (presentations and dashboards), Narrative (analysis and citations), and Model (datasets and Python scripts). Users can upload data and ask AI to generate analysis and narratives, while also viewing the underlying scripts to verify the results. Patel said he hopes to redesign the file itself into a single “.ndoc.html” format that can be opened in any browser, and plans to open-source the format (Apache 2.0) to avoid lock-in by a single vendor.
As of the report date, Ndoc is conducting enterprise pilots and opening its alpha version to a limited number of customers on the waitlist. Registration is free, with usage- and token-based fees; a beta is planned for release within weeks, with a full launch next year. The team currently has three people and has not raised funding. Patel said he wants proof that users love the product before seeking financing, and said part of the reason for leaving was that it takes “hundreds of people’s opinions” for large companies to reach consensus on an open-source format, whereas he wants to move at a speed that AI can enable.
Original sources:
Title: Meta AI director Rohit Patel leaves to take on Microsoft Office and Google Workspace Publisher: TechCrunch (Rebecca Bellan) Publication date: 2026-10-27 URL: https://techcrunch.com/2026/10/27/meta-ai-director-rohit-patel-leaves-to-take-on-microsoft-office-and-google-workspace/ Source nature: media report (exclusive) Verification status: full text read in this instance Supported content: departure, eight-year tenure, MSL director role, Ndoc positioning and product design, pilot status, team size, financing status, and Patel-related statements
Title: Rohit Patel personal website Publisher: Rohit Patel Publication date: unknown (page includes a link to the TechCrunch report dated 2026-10-27) URL: https://www.rohit-patel.com/ Source nature: first-party post/website Verification status: body of the page read in this instance Supported content: founder identity, Ndoc description, and intent to open-source the format (does not prove the company has officially completed the departure announcement)
ABAB AI Insight
Rohit Patel’s work at Meta focused on reinforcement learning and evaluating the trustworthiness of model outputs, and he was involved in research related to the Llama series, including coauthoring the Llama 3 paper with the team and multiple benchmarks. He also previously served on BayPine’s digital advisory board and wrote technical articles explaining LLMs and reinforcement learning. After leaving, he chose to found a product company directly rather than join another large lab, partly because he wants to advance an open file format independently and avoid the multi-round consensus process that large companies use internally for format design.
Ndoc’s path is to build a unified workspace from scratch rather than layering AI on top of existing Office or Workspace products. On funding, the team currently has three people and has not raised capital, adopting a strategy of first validating that users love the product and then seeking financing; its revenue model charges by usage and tokens. This contrasts with OpenAI’s recent ChatGPT in-space workspace and the approaches of major model vendors that connect AI to Microsoft 365 and Google Workspace — the latter more often “embed AI in existing tools,” while Patel explicitly criticizes this approach as “putting lights on a cow instead of building a tractor.”
Recently, Meta Superintelligence Labs and related teams have seen several high-profile departures, including Chief Global Affairs Officer Joel Kaplan, AI researcher Andrew Tulloch, and Emily Dalton Smith, head of AI-for-Work products. Patel’s departure is part of the same window of talent outflow. Compared with historical cases of people leaving big tech to challenge productivity suites, Ndoc’s similarity lies in targeting document, spreadsheet, and collaboration entry points; its difference is that it centers on an open, portable format (.ndoc.html) and verifiable underlying scripts rather than a closed ecosystem.
For Meta, this is the departure of a single individual, and the materials do not show a direct impact on the company’s model development or capital expenditures. For Ndoc, the challenge lies in changing enterprise users’ path dependence on Office and Workspace while competing as a three-person team in a market where OpenAI, Anthropic, Google, and others have already launched enterprise AI tools. Metrics to watch include the conversion rate of enterprise pilots after beta launch and whether the open format attracts third-party developers to build compatible applications.
ABAB News · Cognition Law
A large-tech AI director turning to independently found a productivity tool reflects talent flowing from model labs toward office entry points.
Challenging closed suites with an open, verifiable format creates a mechanism difference from the path of layering AI on existing tools.