MIT Sloan innovation startups pursue AI at scale

Modzy and Snowplow are among the early-stage companies aiming to move AI from science project to enterprise asset. Success will let businesses reap the benefits of the technology.

Enterprises often struggle with getting AI projects out of the corporate skunk works and into widespread use.

The objective, AI at scale, has become a rallying point for technologists in the data science field -- and a key theme among the MIT Sloan CIO Symposium's Innovation Showcase finalists. Many of the early-stage companies selected for the showcase pursue some variation on getting AI and machine learning into the mainstream. They'll be discussing their offerings at MIT Sloan's annual event, which runs May 22-23 in Cambridge, Mass.

Breaking AI out of the pilot stage will move businesses closer to the promised benefits of higher productivity, improved customer experience and novel business models. Enterprises view scaling the technology as critical for achieving their business strategies.

ModelOps: Managing AI at scale

One finalist is Modzy, a Vienna, Va., company that offers a ModelOps platform for deploying and managing AI at scale. The company aims to help customers generate value from their AI spending.

Organizations are investing a lot in data science and analytics today, and not necessarily getting the return from the investment just yet.
Kirsten LloydCo-founder and head of go-to-market, Modzy

"Organizations are investing a lot in data science and analytics today, and not necessarily getting the return from the investment just yet," said Kirsten Lloyd, co-founder and head of go-to-market at Modzy. "Innovation stops in a lab."

Lloyd said she sees ModelOps as providing the same sort of rigor to AI models that DevOps brought to software development. Data scientists can tap Modzy's platform to more quickly integrate AI into enterprise applications, and then track, monitor and secure the models in production to ensure governance, she added.

From the CIO's point of view, Modzy offers centralized management of all the models within an enterprise, Lloyd said. That's an important consideration since data science organizations can be matrixed, with teams reporting to different leaders. As a result, CIOs might lack a sure grasp of AI developments across an organization.

"They can't necessarily track models being built and how they perform over time," Lloyd said.

A ModelOps platform, however, reduces risk while increasing visibility into projects, she said. Cost management is another plus. Centralized management encourages the sharing of AI models, avoiding redundancy and making better use of data scientists' time.

Capturing behavioral data

Another finalist, Snowplow, offers a behavioral data platform that aims to push AI models out of the experimentation zone. The company, which has offices in London and Boston, captures behavior data as users interact with websites, mobile apps and other digital channels. The platform sends that information to a data warehouse, from which enterprises can pursue various use cases.

Seven important benefits of AI for business
Deploying AI at scale could unlock several business benefits.

Snowplow's approach treats the data warehouse as the center of gravity for all things data. "As a data scientist, or data engineer, you treat the data warehouse as the center of your truth," said Nick King, go-to-market specialist at Snowplow.

Centralization contrasts with merging data from a hodgepodge of sources. King described that method as "brittle and cumbersome."

The common source of truth means data scientists can spend more time building models. "These teams that would otherwise have to spend 80% of their time wrangling data [can] consume the data directly from this known and trusted source," said Yali Sassoon, co-founder and chief strategy officer at Snowplow. "They can just focus on, 'What is the algorithm I need to bring to this data set?'"

A trusted data source leads to greater reliability, King noted. That is, data scientists using a consistent approach for training AI models -- and a single, consistent data source -- can expect predictable results with every project.

"The next wave of AI should be more predictive as well as more repeatable," King said. "That increases confidence, and as a CIO, that's the main thing. Can you deliver the same solutions and the same consistency across the organization to maintain the trust of those systems? That actually helps dramatically shift more AI use cases from experimental to applied."

AI abounds

Other AI-oriented Innovation Showcase finalists include Dexai Robotics, a Boston company that features an automated sous-chef built on robotics and AI, and Attestiv, a Natick, Mass., startup that uses AI and blockchain technology to authenticate digital media.

A complete list of finalists is available on the MIT Sloan CIO Symposium site.

Dig Deeper on IT applications, infrastructure and operations

Cloud Computing
Mobile Computing
Data Center
Sustainability
and ESG
Close