One of the biggest pain-points in real-world machine learning is to consistently and rapidly putting models into production. When LeadCrunch[ai] approached Verta, their data science team was only able to deploy about two models annually. After successfully implementing Verta, LeadCrunch sped up model production 5X and now deploys models monthly. Here's how they did it.
LeadCrunch uses Artificial Intelligence and a patented approach to B2B data to identify prospects most likely to become lasting, lucrative clients for B2B sales. It does for business marketers what Facebook does for consumer marketers: finding an audience that looks just like a company’s best customers.
There are many successful players creating lookalike audiences in the B2C space, but B2B lookalike audiences require a lot more research to generate the data that fuels these recommendations.The product is built on complex AI, tens of thousands of data points, and hundreds of models.
Implementing the recommendations from this research can yield impressive results – a typical customer can expect 300% ROI. LeadCrunch serves more than 300 leading enterprises around the world, and its customers benefit by getting actionable intelligence that fundamentally empowers their growth engine.
"Verta.ai freed up a lot of time for my team members to actually spend more time doing exploratory modeling as opposed to having to deal with the heavy engineering aspect of maintaining models."
- Alex Quintero, Manager, Data Science Manager at LeadCrunch[ai]
It takes a lot of data and resources to come up with these B2B audience recommendations. “The complexity involved in distilling these incredible amounts of data into a single data point is among the worst of the problems I’ve faced in my career,” said Steve Biafore, Chief Science Officer at LeadCrunch. When you’re thinking about big things and trying to solve tough problems, it’s easy to make costly mistakes.
LeadCrunch knew that it needed to become more efficient to be competitive. Even with a data science team of eight people, “we realized we were being very slow with delivering value to our customers using our models,” said Alex Quintero, Data Science Manager at LeadCrunch. It was clear that things needed to change.
LeadCrunch needed a solution to make it more productive but it didn’t have the resources or the time to build one. So they turned to Verta, a tool that allows ML practitioners to rapidly version, deploy, and monitor enterprise ML models at scale. “What Verta lets us do is focus on the tough problems and leave the tracking and organization to Verta to help us to not make mistakes that delay production,” said Biafore. “When we tested Verta we realized our time to production for a model was greatly shortened,” said Quintero. “It was easier to deploy the model and maintain the model with a lot less overhead from our engineering team.”
Everyone on the data science team saw an immediate benefit to the ML workflow. “Before Verta, we had 21 pain points identified,” said Jennifer Flynn, Senior Data Scientist at LeadCrunch. “Verta addressed twenty of them, so it was very exciting!”
“With Verta, we can test multiple versions of the model at the same time and not be bottlenecked by engineering. We can run multiple experiments at the same time where we weren’t able to before. And we can use any ML framework that we want, SciKit Learn, Pytorch, Tensorflow, Python 2, Python 3 – any framework at all. That really reduces our reliance on devs.”
- Jennifer Flynn, Sr Data Scientist at LeadCrunch[ai]
Implementing Verta was like gaining two additional data engineers on the team. “It freed up a lot of time for my team members to actually spend more time doing exploratory modeling as opposed to having to deal with the heavy engineering aspect of maintaining models,” said Quinetro.
All of this is backed by a hands-on team of ML engineers. “It’s wonderful just talking to other engineers – they know exactly the problem I am experiencing and how to solve aspects of it,” said Flynn. “They are very responsive!” Verta also allows better team collaboration with its heavy focus on UX and UI. “The dashboard is really wonderful in bringing all of our information together into one place in a comprehensive way, where we know exactly how much experimental space we explored and how our results have been tracking,” said Flynn. “But it’s very easy to use, and its ability to add custom visualizations gave us the flexibility we needed.”
Curious to see what Verta can do for you?
Manasi Vartak is the founder and CEO of Verta, an MIT-spinoff building software to enable production machine learning. Verta grew out of Manasi’s Ph.D. work at MIT CSAIL on ModelDB. Manasi previously worked on deep learning as part of the feed-ranking team at Twitter and dynamic ad-targeting at Google. Manasi is passionate about building intuitive data tools like ModelDB and SeeDB, helping companies become AI-first, and figuring out how data scientists and the organizations they support can be more effective. She got her undergraduate degrees in computer science and mathematics from WPI.
Verta builds software for the full ML model lifecycle starting with model versioning, to model deployment and monitoring, all tied together with collaboration capabilities so your AI & ML teams can move fast without breaking things. We are a spin-out of MIT CSAIL where we built ModelDB, one of the first open-source model management systems.