Very’s Data Science Practice Lead, Jenn Gamble, Ph.D., was recently interviewed on the Human-Centered Artificial Intelligence Podcast, hosted by the DEUS initiative.
Jenn spoke with DEUS Strategic Lead Nathalie Post about why building machine learning solutions is so challenging and the practices we follow at Very to deliver the most compelling results.
Why Organizations Face Challenges Deploying Machine Learning Systems
“Often when we think of machine learning systems and places that do this really well, we think of the big tech companies,” Jenn says. “They often have very large teams and lots of people to be able to do not only the ML model development but also the pipelines, DevOps, ML ops, and ML infrastructure.”
Meanwhile, she notes, countless mid-market and enterprise companies are looking to expand their capabilities in data-driven decision-making, but they don’t have the same well-oiled machines in place that Apple, Microsoft, Amazon, Alphabet, Facebook, and more can access.
To address this gap, Jenn’s been working to build new frameworks that approach the data science development lifecycle from an Agile viewpoint.
Deconstructing Silos in Machine Learning Development
Key to the Agile approach for machine learning development is breaking down silos between the multiple disciplines involved in creating a machine learning solution.
“There may exist some type of data scientist unicorns who can do the majority of what’s required by themselves, but they’re not the typical kind of data scientist you would find in the industry right now,” Jenn says.
She emphasized the importance of partnerships between data scientists and multiple team members with different skill sets, including:
- Software and data engineers, who can help data scientists ask the right questions about the correct abstractions to use, modules to compose to build a pipeline, the deployment process, and more.
- Product designers/front-end developers, who can help data scientists think through what users need and expect in an application and which user flows to employ.
Jenn also highlighted the importance of the tech lead role in every project to direct these multidisciplinary teams. The tech lead focuses on the overall architecture of the system, the different components that need to fit together, how interfaces are specified, the feedback mechanisms, and more. This helps the team to pair the understanding of the product requirements with the technical architecture.
After Jenn and Nathalie’s conversation, one thing is clear: complex machine learning projects require robust, experienced teams. If you’re looking for a team like this to deliver on your next strategic initiative, reach out to our team today.