Despite recent cost cutting and sluggish global economic growth, companies continue to invest in AI products and supporting infrastructure. The message is clear — AI is no longer a competitive advantage, it’s existential.

Yet to support these fast-paced AI initiatives, enterprises have to rely on decades-old data management practices — designed to allow centralized data teams to extract data from sources, push it through pipelines, store it centrally, and only afterward, layer on data cataloging, mastering, semantics, quality checks and access control, all before data becomes usable.

We have talked to well over a hundred companies and one thing is clear: Today’s approach to enterprise data management is too slow, too costly and mostly untrusted. Data teams are overburdened, stitching together fragmented tools, managing complicated and brittle data pipelines, struggling to keep catalogs in sync with reality and stuck in a perpetual cycle of replatforming.

We believe that enterprise data management needs to be reimagined, end to end, with an operating model that is designed for a world where customer expectations change overnight, new competitors emerge from anywhere, and AI agents are everywhere—making real-time, high-stakes decisions at scale. 

Today, I’m thrilled to launch Nextdata OS — the industry’s first unified development and operating system for autonomous data products.

Built from the ground up, Nextdata OS is designed to:

• Simplify the complexities of data management
  
• Speed up AI and application innovation
  
• Scale out safely and efficiently across the enterprise

With Nextdata OS, we’re not launching just another platform — we’re delivering a simpler, smarter way to manage data.

We’ve reimagined data as autonomous, decentralized, self-governing products that work with your existing stack — and work for you.

It’s how enterprises finally scale-out self-serve data with speed, trust, and built-in safety across heterogeneous data stacks.

How we designed Nextdata OS

In designing¹ Nextdata OS our goal was to significantly improve three key metrics for companies. The first: speed. Accelerate the entire data supply chain— from data generation to consumption. We built the system to deliver that speed at scale across large and complex enterprises, and do so with safety and controls built-in.

I knew for any solution to work it had to:

• Embrace the real-world complexity of data — data anywhere, in any shape, and for any use.
• Be decentralized — eliminate any centralized bottleneck in technology, process or team.
• Be radically simple — remove the accidental complexity of years of accreted traditional data management.
• Be open for extension — let users adapt, integrate, build out, and avoid vendor lock-in.
• Be automated and AI-powered — able to bootstrap itself from users’ existing data assets, including ETL pipelines and datasets in warehouses and lakes.

We knew we didn’t need to reinvent everything. Nextdata OS had to work with what users already have  — their existing data compute and storage platforms, as well as their security and quality tools.

Lastly, it was inspired by my work on data mesh and data products — and everything I’ve learned helping large companies transform. We built the technology to let decentralized teams fully manage their own data products, end to end— data products that work for their users, and govern and orchestrate themselves through computation.

The result: the concept of autonomous data products was born, with Nextdata OS as a platform to build and manage them.

What is an autonomous data product?

With Nextdata OS, we’re introducing an upgrade to the concept of data product: autonomous data products.

Unlike first generation data products, Nextdata’s autonomous data products are long-running applications that encapsulate and fully automate the entire data supply chain. Autonomous data products are:

Self-provisioning: Automatically manage their own storage, compute, and other required infrastructure.

Self-publishing: Instantly publish their metadata, health and compliance metrics, access requests, through a uniquely addressable URL — all in near-real-time.

Self-orchestrating: They handle data sensing, generation and access — all within the data product itself, without the need for an external orchestrator.

Self-governing: Continuously validate and enforce policies-as-code throughout their lifecycle. They evaluate data contracts both before consuming upstream data and before sharing it downstream. 

Adapt to any data stack: A pluggable, extensible driver architecture allows them to work with any underlying compute, storage, and quality stack.

Multi-modal output: The same governed semantic can serve multiple formats — a table for analytics, a vector index for agents, or free text for language models.

Domain semantic first: Created by business domains, for business teams — they capture the core meaning and context of the domain from the start.

Their evolution builds on my original work on data products — now designed to work smarter, safer and at scale.

To power autonomous data products, we’ve developed a unique data product containerization technology — and in the coming months, we will share more about how it’s designed and built.

We had to completely rethink what a data product is. Instead of passive files or tables that require constant effort to manage, we built data products that manage themselves.

This shift is key to simplifying data management.

Our goal is to help teams move away from traditional manual data management tasks —

This means no more:

• Building pipelines
• Cataloging data
• Granting access at every storage layer
• Communicating business data through schemas
• Layering on semantics and lineage after the fact
• Catching quality issues only after they’ve caused problems
• Handling every new use case by building a new pipeline
  

Instead, just ship autonomous data products — and let them handle the rest, themselves.

Built in partnership with leading enterprises — and for any team tackling data complexity

From day one, we’ve partnered with some of the world’s largest and most complex companies. We worked with fast-moving business units that need autonomy, accountability, and interoperability with the rest of the organization. We’ve been fortunate to collaborate with visionary leaders driving real change.

A growing number of these companies are adopting data decentralization strategies like data mesh. Their data platform teams — often embedded within business units — are focused on enabling domain teams to build standardized, interoperable, and fully governed data products quickly.

Nextdata OS supports this shift by enabling fast, responsible self-service data product development, while simplifying infrastructure and reducing cost. It does this by eliminating unnecessary intermediate layers found in legacy pipeline architectures like Medallion or Data Vault.

Nextdata OS is already in use across global enterprises — helping teams move faster, lower costs, and increase trust in their data.

Bottom line: scale data products for AI agents and humans — safely and fast

Nextdata OS lets independent teams build, share, discover and manage autonomous data products across organizational and technology silos.

Here are some of the additional capabilities that Nextdata OS offers:

• Effortless data product creation: We respect data product developers’ choice of tools — whether it’s notebooks, SQL dev tools, or IDEs. Nextdata OS enhances their workflow with a standard declarative interface and automation to package and launch autonomous data products. They can use the Nextdata Python DSL, YAML, Nextdata Studio UI, or the Nexty agent for a conversational experience.
• AI-assisted bootstrapping: We make it easy to build your mesh starting from existing ETL pipelines and data. Our Nexty agent can generate your v0.1 mesh in just a few minutes.
• Discover and access: Nextdata OS offers an intuitive — and yes, beautiful :) — way to find, understand, and access data products. It works seamlessly for both humans and agentic apps through APIs. As data products connect, lineage and semantic graphs automatically form, giving you built-in context. And with full-text semantic search, you can easily explore all related data products across the mesh.
  
• Insights at scale: Enterprise-wide dashboards give platform teams real-time visibility into the health, usage, cost, and compliance of data products — across your entire data estate, no matter what tech stack you use.

Built for generative AI — and what’s next

Nextdata OS and its autonomous data products are launching at a time when the data and AI infrastructure landscape is facing a cambrian explosion of tools, architectures and approaches. Everyone is racing to build agents, provide them with real-time context, and make it all work.

We built autonomous data products with this rapid change and uncertainty in mind. They’re multi-modal by design. Thanks to that flexible design, we quickly added support for Retrieval-Augmented Generation (RAG)² and Model Context Protocol (MCP).

Thanks to a simple driver interface, we can rapidly plug in new types of storage and compute, delegating the heavy lifting to the underlying infrastructure. These are just a few examples of the deep design decisions we’ve made to rethink data management for what’s here — and what’s coming.

Join us

I couldn’t be prouder of what our team has created — and even more excited about what it makes possible.

If you’re a data leader, developer, or platform engineer tired of the endless migration cycles and ready for a new way forward, I’d love for you to join us. We are hosting a demo of Nextdata OS at 8 AM PDT on April 22nd and I can’t wait to show you what we’ve built.

👉 Register here to reserve your spot.

It’s time to rethink data management — not as something we constantly fix and rebuild, but as autonomous products that work for us.

Thank you for being part of this journey.

We’re just getting started.

– Zhamak

‍
‍

¹ See The five S’s of Nextdata on our product design values.
² See our webinar on MeshRAG with Nextdata OS.