Rohit Patel, who has spent eight years working on AI at Meta, recently left the company to launch his own startup called Ndoc. As a former director at Meta Superintelligence Labs, Patel brings deep expertise in reinforcement learning and AI evaluation to his new venture. Ndoc is positioned as an innovative, AI-native alternative to established productivity suites like Microsoft Office and Google Workspace, aiming to unify many fragmented work tools into a single platform accessible by entire teams, even if they haven't signed up for the service.

Ndoc’s interface combines three key components: Stage, where users can create presentations and dashboards; Narrative, which offers detailed analyses with citations and generated visuals; and Model, where the underlying data and Python scripts reside. Patel emphasizes transparency by making it possible for users to inspect and understand the AI-generated content at the script level. This feature reflects his belief that AI tools supporting work should be thoroughly auditable, even for users with limited programming knowledge.

Built from scratch using a mix of open-source AI models, Ndoc offers cost advantages compared to solutions relying on frontier models. Patel’s ambition extends to reinventing the concept of the document itself, with a novel ".ndoc.html" export format that can run in any browser and will be open-sourced, inviting other developers to create compatible apps. This open-source approach was a major factor in his decision to leave Meta, where such a project would have faced significant internal hurdles, allowing him to move fast as a solo founder with a team of three.

Ndoc is currently in an enterprise pilot phase, offering early access to a limited alpha user base through a waitlist. Patel plans a beta release soon and a full launch next year, with a pay-as-you-go pricing model. While Ndoc faces competition from AI-enhanced versions of traditional suites being developed by OpenAI, Google, Meta, and others, Patel distinguishes his approach by focusing on creating a fundamentally new workflow platform rather than simply adding AI features to existing tools.