Rohit Patel is leaving Meta after eight years to build Ndoc, a new workspace for company work. The startup aims to bring analysis, presentations and the data and code behind them together, rather than adding AI features to familiar office software.
One workspace for the work behind a decision
Ndoc is designed to gather tasks that can otherwise be spread across Excel, Word, PowerPoint, Tableau, Replit and “ChatGPT copy-paste.” Patel’s idea is a shared space colleagues can access even if they have not signed up for Ndoc.
The product brings three areas into one layout. Stage holds the work people present, such as slide decks, dashboards, simulations or mini apps. Narrative contains the analysis, numbered citations and generated figures that explain the findings. Model holds the datasets and Python scripts behind the output.
That structure makes the result and the work that produced it part of the same workspace. A team can review a presentation, read its supporting analysis and inspect the underlying material without treating each as a separate file or tool.
Making AI work easier to check
Patel’s work at Meta focused on reinforcement learning and evaluating whether AI output is trustworthy. He sees that concern as central to workplace AI: if a system builds spreadsheets and slide decks used for decisions, people should be able to examine how it reached its results.
Ndoc lets users find the Python scripts used to create documents. Patel says people who do not know Python can ask the AI chat to explain a script or identify the line containing a formula. The aim is to make the process inspectable, rather than leave the AI’s work hidden behind the finished document.
Users can upload a dataset and ask Ndoc to analyze it and write a narrative about what it shows. They can also give a document context such as its audience and tone. Patel says the system keeps memory across a person, a particular document, a team and a company.
A different idea of the document
Patel’s larger plan is to rethink the file itself. Ndoc documents can be exported as a single “.ndoc.html” file and opened in any browser. He plans to open-source the format, with the aim of letting other developers build apps that work with it.
That ambition is part of why Patel chose to leave Meta and work independently. He said reaching agreement on an open format inside a large company would involve “hundreds of people and opinions on how the format should look.” Going alone, he said, lets him move at the pace of what AI can do.
Ndoc is built and fine-tuned with a mixture of open source models. Patel says building the platform from scratch makes it considerably less expensive to run than alternatives built on frontier AI models. But early trial users have been more excited about having their work together in one place than about the operating cost, he said.
Winning users will mean changing how they work
Ndoc enters a field where AI companies are building tools that connect to Google Workspace, Microsoft 365 and other workplace apps. OpenAI, Anthropic, Meta, SpaceXAI and Google have released enterprise AI tools for working with those services. OpenAI also launched Space, a shared workspace inside ChatGPT that Patel acknowledges could be a threat.
Patel argues that attaching AI to existing spreadsheets is too limited: he wants a new way to organize work, rather than a tool that fills in cells. Still, replacing established habits is difficult. Ndoc has connectors to import data from tools customers already use, giving teams a way to bring existing material into the workspace.
Patel says he wants to use the product himself before releasing it, and wants proof that customers love it before raising funds. The company has a team of three and no funding. As of Tuesday, it was running an enterprise pilot and opening an alpha to a limited number of customers on its wait list. Signing up is free, but customers will pay for usage and tokens. Patel hopes to release a beta in the coming weeks, with a full launch planned for next year.
The test for Ndoc is whether a unified workspace can persuade people to move beyond the tools they already know. Patel’s approach is to start with a product he wants to use, learn from early customers and keep building quickly. The company is betting that bringing the finished work together with its explanation, data and code will give teams a reason to make that change.