From the Founder: The Wallaroo Community Edition is Here to Get Your ML Ideas Live

I’ve had over 20 years of experience in bringing data and algorithmic products to market. I helped to build the algorithmic and high-frequency trading business at Merrill Lynch from scratch as part of a product innovation team. It was harder than it should have been. While we had organizational, regulatory, data, and legacy technology hurdles to overcome, even once we had overcome those, and had good data and a promising algorithm (“trading strategy”), there was still a lot of work to be done to go from the lab environment to production. This included deployment, scaling, massive infrastructure bills, and ongoing monitoring to ensure our algorithms were helping our bottom line — not losing us money.

While large Investment banks were early adopters of machine learning, over the last five years the adoption of algorithmic techniques in many industries and use cases has exploded. I started Wallaroo to help solve for the “last-mile of production” challenges across any industry. Wallaroo’s mission has always been to enable organizations of all sizes to deliver business value from their data by lowering the barrier of entry to get data algorithms into production (which is where the value creation happens).

As part of this mission, I want our product used by as many users and developers as possible via our free community edition. I firmly believe in the power of the community and the value it brings from a broad cross section of skills and experiences. This includes enthusiasts and early career people discovering new ways to solve problems and build their skills; data teams in small companies and startups deploying a handful of models; and data scientists and ML Engineers from large enterprises hoping to solve complex problems at industrial scale.

In fact, the community connection is an integral part of the culture I want at Wallaroo — a community that is a welcoming place to build experience, exchange ideas, give honest feedback on what works and what to improve, and have a two-way connection with our product engineering and documentation teams.

Our initial foray into community support was to provide a downloadable framework for developers to easily deploy and manage analytics applications without having to worry about infrastructure complexity, scale, or cost. They would just plug in their analytics algorithms into our framework. Their code could be scale and infrastructure agnostic as Wallaroo would take care of the rest while providing real-time monitoring and easy scaling. We started by supporting Python and C++ algorithms.

Sure enough, what we heard from early adopters was that while we had solved many of the challenges around getting analytics algorithms to production and scaling them, there was still a lot more that could be done to help all types of users drive business value. We took that feedback to heart and we doubled down on our mission, especially the “lower the barrier of entry” and “deliver business value” part. We also decided to focus on machine learning algorithms.

And now we feel really good about taking this product, sharpened by further data scientist and engineer feedback from our customers, and releasing it under a free community license. What’s unique about Wallaroo is its powerful combination of ease-of-use, raw compute performance, and full observability that encompasses everything AI teams need to deploy, observe, and manage their ML in production at any scale. It provides:

• A self-service UI and Python SDK for model deployment, management, and A/B testing
• A rust-lang based distributed computing engine that runs on-prem, in any cloud, and at the edge with the super high performance
• Full auditability, data and model validation checks and customizable drift detection

What’s more, Wallaroo provides the flexibility for AI teams to work in any data ecosystem while leveraging the investments they have already made to accelerate time-to-value. Wallaroo works great alongside solutions like Databricks, AWS SageMaker or Google Vertex. Our community edition differs only from the enterprise edition based on limits on scale — all the features are available.

We think our users — data scientists and ML engineers — are going to love Wallaroo. It will empower you to create business value, faster and more effectively — and be bolder with your ideas. Even if you think it needs more work, we want to hear from you. And if you’re struggling to get it working for your use case, we’re here to help — please reach out. Often, that’s a use case that we need better documentation around. Your input will help us build the best product anywhere for any size AI team to easily operationalize ML. I am sure of it.

A short note about who we are.

The most important asset of a company is its culture and its people. We’ve spent the last two years building an amazing team and culture that will do amazing things. My background is in high-frequency algo trading (all ML driven); our VP of Ops came from Qubole (from the founders of Hive and a Databricks competitor); our head of product built the AI platform for Tempus Labs; our VP of Data Science has a PhD in Robotics from Carnegie Mellon University and has helped numerous F100s deliver on AI; and our VP of Engineering comes from DataRobot where he helped scale out their cloud capability. Importantly, we are all focused on the customer experience. There’s a lot still to be done, and with your help we will get it done.