Join top executives in San Francisco on July 11-12, to hear how leaders are integrating and optimizing AI investments for success. Learn More

Enterprises are increasingly looking to AI for opportunities to boost revenue as their operations move online. According to a 2021 PricewaterhouseCoopers survey, a quarter of companies report widespread adoption of AI in their organizations — up from 18% in 2020. But AI projects are at risk of stalling due to the many roadblocks businesses encounter on the pathway to implementation. In a Gartner report, analysts estimate that 85% of AI projects will deliver erroneous outcomes — whether due to bias in the data and algorithms or the teams managing them.

An emerging discipline called machine learning operations, or MLops, aims to prevent these failures by combining machine learning, devops, and data engineering to facilitate the deployment and maintenance of AI models. MLops is a fast-expanding category of companies, anticipated to generate as much as $4 billion in revenue by 2025. Among the players is Wallaroo, which offers a New York-based AI model management platform that can plug into existing systems. Underlining the segment’s growth, Wallaroo today announced that it raised $25 million in a series A round led by M12, Microsoft’s venture arm, with participation from Boldstart Ventures, Contour Venture Partners, Eniac Ventures, and Greycroft, bringing the company’s total raised to $30 million.

Bringing AI to the enterprise

Founded in 2014 as Sendence by CEO Vid Jain, Wallaroo offers services designed to help customers deploy and scale AI their investments. Jain created the concept while working at Merril Lynch, where he realized that the company could derive more value from AI by adopting an improved “last-mile” deployment strategy.

“Data is everywhere, and enterprises across every sector are turning to machine learning to use that data to become more competitive, agile, and profitable. These enterprises, however, are confronting a fundamental roadblock: how to put their machine learning models into production so that those models actually have an impact on the bottom line,” Jain told VentureBeat via email. “Operationalizing data applications at scale is really hard. Existing approaches — whether containerization, cobbling together various existing technologies, or customizing an analytics workhorse like Apache Spark — are cumbersome, limited in scope, expensive at scale and prone to failure.”


Transform 2023

Join us in San Francisco on July 11-12, where top executives will share how they have integrated and optimized AI investments for success and avoided common pitfalls.


Register Now

The Wallaroo platform is built on four components: MLops, a distributed processing engine, data connectors, and audit and performance metrics. It’s Jain’s assertion that by combining these, Wallaroo can run multiple models on shared infrastructure without adding significant overhead to the models’ compute times.

With Wallaroo, data scientists can deploy models against live data to testing, staging, and production environments using popular machine learning frameworks. The data processing engine, which compiles to native code and can run in the cloud, on-premises, or at the edge, provides a place to evaluate production models as well as perform pre- and post-processing tasks. Wallaroo’s data connectors enable integrations with enterprise data sources as well as custom integrations with in-house solutions. And the observability tools deliver audit logs, compute and model performance metrics, and A/B testing comparisons to data scientists, operations, compliance and risk teams, business heads and finance leads, and other stakeholders.

A diagram showing the Wallaroo platform’s implementation.

“From the beginning, our team’s mission has been to build from the ground up a new way to deploy computational algorithms with breakthrough performance and scalability and to work in any cloud, on-premises, or at the edge — and without having to rip out their existing tools in the process,” Jain added. “[We plan to] release a free community edition to put Wallaroo into the hands of more data scientists and machine learning engineers, generating the ongoing feedback needed for rapid product improvements, as well as building a network of product champions. [We also plan to] make Wallaroo usable across a wider range of machine learning workflows and model development environments, and to provide more of out-of-the-box connectors to other data and analytics products and services.”

MLOps on the upswing

Even with MLops, not all enterprises are realizing the benefits of AI. The majority of data scientists say that only around 20% of models generated to be deployed have reached that point, and MIT Sloan Management reports that a measly 10% of companies obtain “significant financial benefits” from AI.

“MLops may be a key part of [AI project] fulfillment, but it isn’t a deployment cure-all,” wrote Eric Siegal for KDNuggets earlier this month. “The greatest bottleneck for deployment is usually gaining buy-in from human decision makers, even if the integration challenges are also impressive … If it turns out that MLops is the main missing ingredient, that’s due to a failure of leadership, not an argument that MLops is the singular solution to our deployment woes in general.”

Wallaroo claims that it’s beating the odds, however, with customers including a Fortune 100 cybersecurity firm and companies in real estate, manufacturing, and adtech. Controversially, Wallaroo has also worked with the U.S. Military to analyze data from internet of things devices, such as drones and ships, to detect security anomalies. We’ve reached out to the company for clarification on its defense contracts and will update this article once we hear back.

To maintain pace with competitors like Comet, Iterative, ZenML, Domino Data Labs, and Landing AI, 24-employee Wallaroo says it’ll invest the funds from the latest round into product development and hiring.

“Enterprises have become more serious about actually getting to business value from their AI investments. Accelerated digital transformations in industries, forcing them to start getting more serious about AI because they are seeing the need for (e.g., more resilient supply chains) — or to act sooner on massive changes in the market or better personalize the consumer experience,” Jain continued. “[Illustrating the trends, our revenue is] in the millions with triple-digit growth in the next few years based on current pipeline. [We have] customers across industries — finance, adtech, retail, manufacturing, public sector — for all sorts of use cases including dynamic pricing, cybersecurity, internet of things, marketing optimization, and more.”

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.