After spending eight years developing artificial intelligence systems at Meta, Rohit Patel has decided to launch his own venture. The former director at Meta's Superintelligence Labs told TechCrunch that he has left the company to establish an AI startup called Ndoc. His ambition is to replace everyday productivity tools such as Microsoft Office and Google Workspace.
Patel's exit comes amid a series of high-profile departures from Meta, including Chief Global Affairs Officer Joel Kaplan, AI researcher Andrew Tulloch and Emily Dalton Smith, who led product efforts for AI-for-Work. Patel is now drawing on his experience to develop an AI-native workspace from scratch.
An all-in-one workspace
Rather than making users switch between Excel, Word and PowerPoint or manually transfer content from ChatGPT, Patel wants Ndoc to bring all office tasks together on one platform. The software replaces conventional productivity applications with a straightforward interface divided into three tabs.
The 'Stage' tab contains presentations, dashboards and slide decks. The 'Narrative' tab is dedicated to detailed reports, charts and explanations of the data. The 'Model' tab stores the underlying data and the computer code used to process it.
Drawing on his experience evaluating AI reliability, Patel has designed Ndoc to prioritise transparency. Whenever the AI generates a spreadsheet or chart, it also displays the precise Python script used to produce the calculations. Users without programming knowledge can ask the integrated AI chatbot to explain the code line by line, allowing them to understand how the results were generated instead of relying on calculations hidden behind the interface.
Reinventing the simple file
One of Patel's central ideas is to fundamentally change how digital files work. Ndoc allows users to export their projects as a single file with the extension ".ndoc.html", which can be opened in any standard web browser.
Patel intends to make the file format open-source so other developers can create compatible tools without requiring users to remain within the Ndoc ecosystem.
This commitment to an open file format was a major factor behind his decision to leave Meta. At a large technology company, securing approval for such an initiative could involve hundreds of people. Running his own startup, Patel believes, will allow him to work at the speed of AI development rather than get held back by lengthy corporate approval processes.
Taking on the tech giants
Competing with established companies such as Microsoft, Google and OpenAI is an ambitious undertaking for a startup with just three team members and no external funding. While most major technology companies are adding AI capabilities to their existing document and spreadsheet products, Patel believes a fundamentally different approach is needed.
He likened the current technology landscape to discovering electricity but using it only to attach lights to bulls working in a field, instead of developing a tractor. Ndoc, in his analogy, is intended to be that tractor.
To keep operating costs below those of competitors, the startup uses a combination of open-source AI models instead of relying exclusively on expensive, large-scale frontier models.
Ndoc also includes built-in connectors that allow users to bring in existing information from the productivity tools they already use.
When can you use it?
Patel is delaying plans to seek outside investment until he can demonstrate that customers genuinely value the product. Ndoc is currently in its alpha testing stage, with a small group of early users drawn from a waiting list.
Although users can sign up for the software for free, the company plans to charge customers based on usage in the future. A beta release is expected within the coming weeks, while the full commercial launch is targeted for 2027.
