Meta AI Director Rohit Patel Exits After 8 Years to Launch Ndoc Rival to Microsoft Office and Google Workspace

TL;DR

• Meta AI Director Rohit Patel is leaving Meta after eight years leading AI evaluations and reinforcement learning work to found his own startup.
• His new company, Ndoc, is building AI-native productivity tools designed from the ground up around large language models, not bolted-on copilots.
• Ndoc aims to take on Microsoft Office and Google Workspace with a unified docs, spreadsheets, and collaboration platform focused on automation and verification.

A Veteran of Meta's AI Push Moves On

After eight years at Meta, AI Director Rohit Patel is leaving the company to strike out on his own.

Patel confirmed the move in a post shared this week, saying his last day at Meta will close out a run that spanned some of the most transformative years in artificial intelligence. During his tenure, he led teams focused on AI evaluations and reinforcement learning, two areas that became central to Meta's development of Llama models, AI assistants, and broader efforts to make large language models safer, more capable, and more reliable.

In his farewell note, Patel described the decision as bittersweet, thanking longtime collaborators and pointing to Meta's open-source AI strategy as a highlight of his time there. He said he was proud of the progress his teams made on model evaluation frameworks and RL-based post-training, which have become industry-standard practices for testing reasoning, instruction-following, and factuality.

Sources familiar with the move say Patel had been exploring founder ideas for months before deciding to leave, with encouragement from former Meta colleagues now in the startup and venture world.

What Is Ndoc?

Patel's next chapter is Ndoc, an early-stage startup aiming to rethink productivity software for the AI era.

While details are still emerging, Ndoc is described as an AI-native alternative to Microsoft Office and Google Workspace. Rather than adding an AI assistant on top of traditional documents and spreadsheets, the startup wants to make AI the core interface for creating, editing, analyzing, and collaborating on work.

According to early descriptions, Ndoc will launch with a unified platform covering documents, spreadsheets, and team workspaces. The vision is a system where users can prompt, delegate, and verify — for example, asking Ndoc to draft a report from raw data, build a financial model from a plain-English description, clean up a messy dataset, or summarize and cross-check a 50-page document.

Patel has said Ndoc is built on the idea that current office suites were designed for manual typing, clicking, and formatting, while future knowledge work will be about directing and auditing AI agents.

Why Take On Microsoft and Google Now?

Challenging Microsoft Office and Google Workspace is no small task. The two giants dominate global productivity software, with Microsoft 365 Copilot and Google Gemini deeply embedded across enterprise workflows.

But Patel and a growing wave of AI founders argue that incumbents are constrained by legacy architecture. Both Microsoft and Google have added generative AI to Word, Excel, Docs, and Sheets, yet critics say those features still feel like add-ons rather than fundamental redesigns.

Ndoc is betting that a startup can build faster and think differently, with key differentiators expected to include:

Agent-first workflows that can perform multi-step tasks across docs and spreadsheets autonomously
Built-in verification and citations to reduce hallucinations in business-critical documents
Enterprise-grade privacy controls and permissions for AI accessing sensitive company data
Seamless collaboration that blends human edits, comments, and AI-generated changes

The timing aligns with surging investor interest in AI productivity startups. Companies like Notion, Coda, Gamma, and a crop of AI spreadsheet and doc startups have raised hundreds of millions in the past year, all arguing that the $100 billion-plus office software market is ripe for disruption.

From Evaluations Expert to Founder

Patel's background could give Ndoc an edge on one of AI productivity's biggest problems: trust.

Having spent years leading AI evaluations at Meta, he specialized in measuring model accuracy, bias, reasoning failures, and risks. That experience with reinforcement learning and systematic testing is directly relevant to building AI tools that businesses can rely on for finance, legal, research, and operations.

In interviews and social posts previewing the startup, Patel has emphasized accuracy over flash, saying Ndoc will focus on making AI outputs auditable, reproducible, and controllable. The goal, he said, is not a demo that writes a fun paragraph, but a tool a CFO or lawyer would trust with real work.

He has not yet disclosed funding, headcount, or a launch date, but said Ndoc is hiring across AI research, engineering, and design, with a private beta expected in the coming months.

What's Next

Patel's departure continues a broader trend of senior AI leaders leaving Big Tech labs to found startups. In the past two years, directors and researchers from Meta, OpenAI, Google DeepMind, and Anthropic have launched a new generation of agent, coding, and enterprise AI companies.

Whether Ndoc can truly rival Microsoft Office and Google Workspace will depend on execution, distribution, and enterprise trust — hurdles that have sunk many Office killers before.

But with deep expertise in evaluations and reinforcement learning, and a clear thesis that productivity software needs to be rebuilt from scratch for AI agents, Patel's move is one to watch. If Ndoc delivers on its promise of AI-native docs and spreadsheets that actually get work done, the future of office work could look very different from Word and Docs.

AndroGuider Team

Articles written by the AndroGuider team. We try to make them thorough and informational while being easy to read.