Large tabular models are new to the enterprise AI landscape. Because they're new, we're seeing teams explore different ways to deploy them depending on their infrastructure preferences, operational maturity, and how much control they want over the underlying stack.

At Fundamental, we offer three paths to run NEXUS:

1. Our SaaS option requires zero infrastructure on your end.

2. Self-hosted on EC2 gives you maximum control over the environment.

3. And now, NEXUS runs on Amazon SageMaker, which sits between the two and offers a managed experience without giving up the benefits of running in your own cloud account.

This post focuses on what makes the SageMaker deployment a strong starting point for data science teams that want to move fast without taking on too much operational overhead.

Simpler Setup

Standing up NEXUS on EC2 is straightforward. We provide a CloudFormation template that provisions the instances, configures security groups, sets up networking, and handles the install for you. It's a solid path, but once it's running, your team owns the ongoing care of those services and the underlying infrastructure.

SageMaker takes most of that off your plate. You're working inside a managed environment that already handles the underlying compute, networking, and lifecycle management. Pointing NEXUS at SageMaker is a much shorter path from zero to a running model, and there's less for your team to maintain over time.

If you want to see the install in action, we have a walkthrough video that covers the full setup on SageMaker step by step.

Faster Iteration

Once NEXUS is running, the bigger win shows up in how quickly your team can iterate. Managed infrastructure means you're not waiting on instances to spin up, capacity planning for experiments, or coordinating with platform teams when you need to scale up for a larger training run.

For data scientists, this is the part that matters most. The time between having an idea and testing it shrinks. You spend more time on model evaluation and the actual decisions that make a model useful, and less time on the mechanics of keeping infrastructure healthy.

Our first API call video shows what this iteration loop looks like in practice once you're set up. It's a good reference for understanding the workflow you'd settle into day to day.

Lower Operational Overhead

The third area where SageMaker pulls its weight is ongoing operations. Scaling is handled automatically based on workload. Updates and patches to the underlying infrastructure happen without your team needing to plan maintenance windows. Monitoring is built in, so you have visibility into how your models are performing without standing up a separate observability stack.

For teams that are running NEXUS in production, this matters. You want your data scientists working on models, not babysitting servers. SageMaker reduces the surface area of things that can go wrong operationally, which frees up your team to focus on the parts of the workflow that actually move the business forward.

This doesn't mean EC2 is the wrong choice. Teams with strong platform engineering capabilities or specific compliance requirements often prefer the control that EC2 gives them. And teams that don't want to manage any infrastructure at all are well served by our SaaS offering. The point is that SageMaker fills a real gap between those two ends of the spectrum.

Choosing the Right Path

The right deployment depends on your team's setup. If you want zero infrastructure, our SaaS offering is the fastest way to get going. If you need full control and your team is comfortable managing cloud infrastructure, EC2 is a great fit. And if you want NEXUS running in your own AWS account but don't want to take on the operational burden of self-hosting, SageMaker is the path we'd point you toward.

For most data science teams getting started with tabular foundation models, SageMaker hits a sweet spot. You get the benefits of running inside your own AWS environment, including data residency, security controls, and integration with the rest of your AWS stack, without the operational overhead of managing the underlying infrastructure yourself.

Try It Out

If you're ready to get started, our installation video walks through the setup on SageMaker, and our first API call video shows you how to start running predictions. Both are short and practical, designed to get you from interested to productive as quickly as possible.

Large tabular models are still a new category, and we're working with customers to figure out the best ways to deploy them in real production environments. SageMaker is one of those paths, and for a lot of teams, it's the right one.