At Dataiku Everyday AI events in Dallas, Toronto, London, Berlin, and Dubai this past fall, we shared a common thread with the audiences: enabling organizations with standout capabilities that make it even easier to make data, analytics, machine learning, AI, and GenAI the source of their competitive differentiation in the years to come. The key to doing so? Moving beyond point solutions to leverage Dataiku, the Universal AI Platform. The link to check out the full video is below, but if you would rather read about the session and its key highlights, continue on!
What Exactly Do We Mean by "Universal"?
When we mention Dataiku is the Universal AI Platform, we mean universal in four main ways:
- First, your people. Dataiku isn’t just a platform just for coders. It isn’t just a code-free tool, either. It’s for everyone. From data scientists and data or ML engineers to business analysts and domain experts, from AI builders to AI consumers, Dataiku meets users where they are. That means coders are first-class citizens in Dataiku, as are business users. The beauty of Dataiku is that everyone is working together in one place.
- We also mean universal when it comes to data. Dataiku supports data of any type, wherever it’s stored — whether it’s on prem, in the cloud, in files/spreadsheets, in your CRM, or a little bit of all of the above. Dataiku works with your data wherever it resides.
- Dataiku is also the Universal AI Platform when it comes to governance. Dataiku introduces governance across the AI lifecycle by design because of our signature, visual “Flow” that allows you to see exactly where data is coming from, how it’s being transformed, and how it’s being used for ML or with generative AI technology. In addition, we also offer Dataiku Govern, a module that allows you to manage your entire AI portfolio against regulatory objectives and AI governance principles.
- And finally, the Universal AI Platform is about your technology. At Dataiku, we like to call ourselves “aggressively agnostic” — this has always been our philosophy. Because we know that the best and hottest technology of today won’t be the best of the best tomorrow. And our customers need to be able to switch technology with ease while keeping the same interface and interaction experience for the people working with data.
Putting It Into Practice
We’re focused on accelerating the AI trajectories of our customers in three key ways:
- Allowing them to build enterprise-grade generative AI. That means doing it quickly but also doing it safely.
- At the same time, we’re allowing our customers to democratize and accelerate across their business. Not just generative AI, but all types of analytics, “traditional” ML, and data work.
- And finally, we’re ensuring that our customers have the comprehensive XOps and AI governance systems in place to build and maintain data products at scale.
We’ll break down these three areas in more detail in the following sections.
1. Enterprise-Grade Generative AI
Let’s talk about what building generative AI applications in the enterprise actually looks like.
First, we have a slew of LLM providers, models, AI services, etc. to choose from. Using one or more of these, you have to then leverage tools, prompting, potentially build knowledge bases that leverage your own data, and combine it all together to either expose some application to end users in some kind of dedicated web application (like a chatbot) or embed this application in an existing tool. From there, this process, this application, needs to be integrated into your larger systems for operationalization, ensuring that the quality is good, that it’s being governed, maintained over time, etc.
In a nutshell: It’s relatively straightforward to build one application. But you can see how building 20, 50, 100 can get messy quickly. This is a familiar story — it’s the same challenge data science projects have always had. At Dataiku, we do believe that enterprises will be building not one or two but potentially hundreds of these applications. That’s why, from the start, we’re thinking about scale, we’re thinking about providing comprehensive GenAI capabilities. How?
Let’s start from the bottom, the base — this is the Dataiku LLM Mesh. We built this common backbone layer to enable choice among a growing number of generative AI services and models. The Dataiku LLM Mesh acts as a secure API gateway to break down hard-coded dependencies and manage and route requests between applications and underlying services. For secure access to hundreds of top LLMs, we partner with leading providers of hosted LLM services as well as Hugging Face for accessing open-source models. The LLM Mesh also handles access to vector databases and accelerated and containerized compute capabilities that allow you to self-host and serve local LLMs easily and efficiently, if that’s your choice.
The LLM Mesh is really the core of our comprehensive GenAI offering. But of course, a secure gateway for LLMs only makes sense if you are able to then build GenAI use cases quickly. With Dataiku, you have all the tools required for builders across the business to bring generative AI applications to life — fast and with or without code per their preference.
From LLM-powered recipes for common tasks like sentiment analysis to Dataiku Prompt Studios, where you can iteratively design and evaluate prompts to Dataiku Answers, a pre-built chat frontend — Dataiku provides GenAI builders with everything they need to create generative AI applications at scale.
That’s great, you might be thinking, but I don’t really want tons of people building hundreds of applications quickly — sounds like it might get out of hand. That’s why the next component is Dataiku LLM Guard Services, a layer that reduces overall risk around cost, potential safety concerns, and quality.
The final layer is generative AI governance. Dataiku offers robust capabilities to ensure that you’re able to keep pace not only with regulatory requirements but also with the governance standards of your organization.
2. Democratization and Acceleration of Data Activities in the Enterprise
Here, we’re talking about speeding up data workers across ALL activities, and empowering more people across the business to use data for their work.
This is something that has always been near and dear to our hearts at Dataiku. From robust AutoML and explainability features to code-free data transformation, we’ve built in democratization and acceleration from the beginning. Even hardcore data scientists often tell us about how they leverage the visual features in Dataiku simply because of how much faster they are able to go. But today, we’d like to dig deeper into a few new ways Dataiku is bringing democratization and acceleration.
The first is through productivity assistants. That’s right, we are using generative AI WITHIN Dataiku to provide an array of assistants to help people — from coders to business users — be more efficient and productive.
The first, and one of the most popular, is AI-enhanced data preparation. In Dataiku, all you have to do is type your data transformation steps in natural language and watch it work, lowering the barrier for everyone to prepare and transform data for their projects. Similarly, for coders, we’ve also rolled out code assistants in both Code Studios and notebooks.
Finally, to help communicate out to stakeholders, we’ve introduced Flow documentation and explanation, which automatically details what data is being used, how it’s being transformed and processed within a flow. We know that documentation is not a task that most people enjoy and as a result it often doesn’t get done, but it’s a critical part of collaboration, communication, and governance, so Dataiku is here to help with that.
3. XOps and AI Governance Readiness
For us, XOps is about bringing ANY kind of data product to production and maintaining it throughout its lifecycle, whether “traditional” ML, GenAI, or simple automation of data pipelines. It’s about addressing the needs of all stakeholders in those processes, removing friction to automate and simplify, and turning arduous processes into repeatable approaches.
We’ve been focused on three capabilities in particular:
- External models: We recognize that our companies often build models in different places and using different tools, and may have models running somewhere else. Dataiku external models are a way to surface, evaluate and use models in Dataiku that are already deployed on Amazon SageMaker, Azure Machine Learning, Google Vertex AI, or Databricks.
- We also recognize that our users want to deploy anywhere, even outside of Dataiku. And we allow for that with Dataiku. This reinforces the democratization we talked about earlier because it especially allows less technical personas, like data and business analysts, to build ML models and deploy them to already established infrastructures, taking stress off ML engineering teams.
- Finally, we’ve been focused on unified monitoring. Let’s face it: Ops people are overburdened because they have multiple models or projects running on different platforms, and they want to make sure that everything stays under control. Dataiku unified monitoring provides out-of-the-box dashboards allowing you to monitor the health and status of Dataiku projects, via Dataiku endpoints and other endpoints not managed in Dataiku. So even with models or projects deployed in different places, the governance of those projects can still happen all in one place — Dataiku.
In addition to Dataiku itself and the way it’s designed being governance-friendly, we also have Dataiku Govern as mentioned earlier, that specifically allows for management of your AI portfolio against regulatory objectives or obligations as well as your own internal AI governance principles.
This includes in particular our EU AI Act Readiness solution, and we will insist on the “readiness” naming as indeed it is not here to guarantee compliance but rather ensure readiness. The team has done a lot of work to fully build out this template that makes mapping against the EU AI Act risk tiers much much more straightforward — a very important solution especially across the EU and for multinational companies in the years to come.
Conclusion
In today’s rapidly evolving AI landscape, standing out requires more than just adopting the latest tools — it demands a universal approach to AI that empowers people, data, and technology. At Dataiku, we’re committed to helping organizations move beyond fragmented solutions to unlock enterprise-grade generative AI, democratize data activities, and ensure robust governance at every step. Whether you’re building your first application or scaling to hundreds, Dataiku provides the foundation for long-term success.