Personify Health Reduces AI Model Development Time by 82% for Personalized Healthcare
Personify Health tailors wellbeing recommendations for millions of members with a single AI platform that unites engineers, data scientists, and analysts for predictive and generative AI.
Before DataRobot, it took three weeks to deliver a new model. Now, we average 21 hours for a viable model end-to-end. It’s powerful to finally have integrated automation that delivers all of this.
Harnessing AI to improve healthcare with personalization
Personify Health is leading the AI revolution in healthcare, delivering personalized health and wellbeing recommendations to their 19 million members and their families.
Over 6,000 employers — including 25% of the Fortune 500 — rely on the company to help their employees make better health decisions and ultimately reduce the cost of care.
“Our mission is to empower people to take control of their wellbeing and healthcare,” said Amit Jain, CTO. “At the end of the day, it is very important that we provide advanced personalization for every individual user and AI is central to that.”
As the first personalized health platform company, Personify has deep expertise in applying AI and machine learning to empower people to live healthier lives. Their team of experts across engineering, data scientists, and analysts had spent years delivering large-scale predictive analytics while adhering to strict data governance practices. Though robust, their massive AI infrastructure was creating costly inefficiencies.
“Integrating and orchestrating hundreds of models at scale is challenging,” Jain said. “With a lot of fragmentation and silos forming over the years, we needed one platform to bring all our efforts together and quickly train and deploy models in parallel.”
AI model development and delivery at scale in the healthcare industry
When looking to replace its legacy AI system, Personify Health sought a platform that could accelerate time to value without sacrificing accuracy. The company also wanted a solution to link its many systems and data sources and keep up with Personify’s fast-paced continuous integration/continuous deployment (CI/CD) cycles.
The company found that DataRobot checked all the boxes. It unifies its predictive and generative AI across teams, technologies, and data sources, therefore accelerating time to production.
“DataRobot shines in bringing all the pieces together for us,” Jain said. “The API-based ecosystem allows us to orchestrate model development and delivery at scale.”
DataRobot accommodates the company’s large, distributed, cloud-native architecture. As an environment-agnostic platform capable of linking to disparate data sources and Personify’s many technology partners, including AWS foundation models from Titan and Claude 3, DataRobot proved to be the missing link for Personify’s years of investment in AI tools and expertise.
DataRobot shines in bringing all the pieces together for us. The API-based ecosystem allows us to orchestrate model development and delivery at scale.
Reducing delivery time by 82% with 80% accuracy
In DataRobot, data scientists can choose from a large library of readily available, qualified algorithms and quickly benchmark to decide if they are the right fit. Personify gathers more than 700 features in most models, and trains them on constantly changing data.
“Producing hundreds of models multiplied by millions of members in a highly accurate fashion is critical for user engagement,” Jain said. “We are proud to be able to do this at scale for every member on our platform today.”
The end-to-end AI process lets them train models and compare outputs to find the highest-performing and most useful models rapidly. That agility accelerates model delivery and allows data scientists to pursue multiple models in parallel.
“Before DataRobot, it took three weeks to deliver a new model,” Jain said. “Now, we average 21 hours for a viable model end-to-end. It’s very powerful to finally have integrated automation that delivers all of this.”
Those time savings ensure that Personify’s team delivers greater value to their customers with less effort and less wasted time. According to Personify Health, DataRobot has also multiplied computing power by 8-10X and supported significant accuracy improvements — more than 80% accuracy on most models.
Moreover, stronger visibility and explainability behind models have increased confidence in their answers and support the company’s extensive governance practices, including regulatory compliance and improved decision-making among team members. It also amplified communication with Personify’s AI executive steering board and governance team who guide every decision. DataRobot aligns with Personify’s rigorous approach by providing model governance and back-end explainability behind every model.
Producing hundreds of models multiplied by millions of members in a highly accurate fashion is critical for user engagement. We are proud to be able to do this at scale for every member on our platform today.
The benefits of using AI to provide personalized health support for millions
Personify pursues a wide range of predictive AI use cases to elevate how it serves customers and their members. Models help them uncover patterns, which enable tailored recommendations for happier, healthier individuals and families. Other AI use cases include reaching out to members via targeted campaigns for fully customized wellbeing.
Already, Personify has centralized its AI efforts and shortcut processes by weeks, supercharging their existing expertise and empowering their data science teams to deliver better support to their customer community. The company looks forward to further accelerating time to value by going deeper with DataRobot.
DataRobot applications provide intuitive insights and show significant promise for further efficiency shifts. The templates offer a decision intelligence feature, empowering repeatable use case logic with components the team can adapt for specific business problems.
Templates also make it easier to rerun processes, share work across the Personify team, and collaboratively develop end-to-end solutions tailored to their specific use-case requirements and stack. The result? Fewer silos, fewer process headaches, and a faster time to production.
“Getting from idea to model implementation has always been a pain point for data scientists,” Jain said. “DataRobot applications dramatically shift the curve for us by reducing the time from months to hours. That’s a huge win and a fundamental change in how we productionalize models.”
Partnering with DataRobot also ensures scalability with generative AI; as the company grows, the number of genAI use cases increases. Jain appreciates that DataRobot can manage and monitor both predictive and generative models in one place.
“DataRobot is a core part of how we build, execute, and deliver,” Jain said. “Our ability to accelerate from several weeks of development to sometimes under a day is tremendous. That’s possible because of DataRobot and our team is super excited to partner with DataRobot.”
DataRobot is a core part of how we build, execute, and deliver. Our ability to accelerate from several weeks of development to under a day is tremendous. And that’s possible because of DataRobot.
