In this episode, we’re joined by Amy Barger, DMI’s Vice President of Digital Business Solutions, and Josh Elliot, COO and Co-Founder of Modzy, a leading provider of an enterprise Machine Learning Operations (MLOps) platform that accelerates and eases the deployment and scaling of production machine learning.
Our guests discuss how Machine Learning Operations can help organizations and agencies quickly deploy and scale AI models in production. They also discuss the challenges organizations face when embracing AI/ML and share advice on how organizations can accelerate their business goals with AI/ML.
GUESTS: Amy Barger, Vice President of Digital Business Solutions, DMI; Josh Elliot, COO and Co-founder of Modzy
HOST: Liv Crowe, SVP of Platforms, DMI
EPISODE HIGHLIGHTS:
7:18 – The Expansion of AI Beyond Experimentation into Production
8:01 – Defining MLOps and Measuring its Value
9:39 – Challenges Faced by Companies Embracing AI
11:40 – Managing Risks
14:38 – Preparing for Resistance with a Holistic Change Management Plan
16:05 – Going Deeper on the Benefits of MLOps
19:20 – Advice for CEOs and CIOs Getting Started on the AI Journey
21:06 – Use Cases for MLOps
24:40 – Adopting MLOps in the Federal Government
EPISODE SHOW NOTES:
In this episode, we’re joined by Amy Barger, DMI’s Vice President of Digital Business Solutions, and Josh Elliot, COO and Co-founder of Modzy. Our guests discuss how Machine Learning Operations can help organizations and agencies quickly deploy and scale AI models. Amy and Josh also discuss the challenges organizations face when embracing AI/ML and share advice on how organizations can accelerate their business goals with AI/ML.
7:18 – The Expansion of AI Beyond Experimentation
A recent Gartner report found that over 80 percent of executives think automation can be applied to any business decision. “Many organizations are starting to expand their use of AI beyond experimentation and running just a few machine learning models. We’re starting to see some of the larger tech players growing to hundreds, thousands, or even millions of models in production,” says Josh.
8:01 – Defining MLOps and Measuring its Value
Josh cites a 2021 McKinsey report that highlights the impact of AI investments that use MLOps practices. “With an MLOps solution, you can absolutely enable faster deployments, better quality control through standards and increase your organization’s ability to rapidly respond to changing business or mission needs,” says Josh.
9:39 – Challenges Faced by Companies Embracing AI
“Almost every organization wants to be an AI enabled company,” says Josh. “[But] we’re seeing teams struggle with the complexity and length of development cycles.” Many organizations implementing AI run into talent gaps, struggle with production implementations, and are overwhelmed by the need to define processes, governance risks and controls.
11:40 – Managing Risks
“There are both intrinsic and extrinsic risks at every turn, but also opportunities to manage those risks,” says Josh. He gives several examples, including impact assessment frameworks that exist to help guide organizations through the risk management process. “Key is understanding the scope of the stakeholders and the magnitude of impact to each of them if something were to go wrong with your machine learning algorithm.” Other risks that need to be managed include quality and limitations in data, model performance drift and regulatory changes that can impact one’s industry.
14:38 – Preparing for Resistance with a Holistic Change Management Plan
Amy points out that another risk for AI/ML adoption is internal resistance. “It’s natural to anticipate that some of these AI initiatives may be met with resistance,” she says. “It’s important to make sure you have a holistic change management plan in place to help these organizations prepare – and, of course, also increase adoption.”
16:05 – Going Deeper on the Benefits of MLOps
Gartner projects that, by 2026, organizations that operationalize AI transparency, trust and security will see their AI models achieve 50 percent improvement in terms of adoption of business goals and internal user acceptance. Key to that improvement, Josh argues, will be the introduction of an MLOps solution. “One of the things that we’re seeing with some of our customers: before introducing an MLOps solution, it was taking them on order of six to nine months to deploy just one model into production. After implementing the standards and the processes associated with an MLOps solution, they’re seeing that reduced down to hours. So that’s a significant efficiency gain for those organizations.” Beyond efficiency, an MLOps platform ensures models run and scale in a consistent way and provides an API-based tool that connects to existing CICD pipelines, data storage solutions and enterprise software.
19:20 – Advice for CEOs and CIOs Getting Started on the AI Journey
“Organizations should be looking for solutions that support composable business and for technical architectures solutions that easily integrate into existing workflows and tech stacks without causing major disruptions or creating additional friction,” says Josh. He adds that MLOps solutions aren’t just for mature organizations. IDC predicts that by 2024 60 percent of all enterprises will have operationalized their ML workflows through MLOps capabilities.
21:06 – Use Cases for MLOps
Josh cites the example of a major industrial vehicle manufacturer that Modzy has partnered with. “They initially started to focus on manufacturing floor safety and quickly realized that once they were able to implement the machine learning ops capability, they were able to expand into other use cases, such as inventory status.” MLOps allowed the manufacturer to “predict delays across their entire supply chain, whether that was due to part identification or mismatch or geopolitical events that could cause significant impact on the production of the raw materials.” Other use cases discussed by Josh, Amy and Liv come from the automotive and telecom industries.
24:40 – Adopting MLOps in the Federal Government
“In the Federal space, we’re really just starting to explore opportunities, as the government is further maturing and scaling AI,” says Amy, citing a case from one of DMI’s Health and Human Services customers that has an automated conversational assistant. “The challenge being faced there is that once these models are deployed, they have no way to track and tell how well they’re working in production. Using an MLOps platform that has these monitoring capabilities built in would allow the customer to identify any performance issues and optimize those models accordingly and, in the long run, improve the quality of their predictions.”
About Modzy:
Modzy is an enterprise-grade model operations platform that provides teams features to deploy, connect, and run AI models anywhere across the enterprise. With APIs and pre-built connectors for development tools and the ability to deploy on-prem, in the cloud, hybrid, or at the edge, Modzy improves teams’ development productivity, speed, and reliability for AI-enabled solutions.
