DataRobot now covers entire generative AI lifecycle for enterprises

Boston-based DataRobot, a unicorn startup that offers a platform for enterprise AI development, is going all in on generative AI support. The company today said it is updating its offering with new generative AI-specific capabilities and applied services to give teams an open and end-to-end solution for experimenting with, building, deploying and monitoring enterprise-grade AI assistants.

The development follows DataRobot’s March update and will make it easier for teams to go from concept to value with gen AI. It comes at a critical time as almost every enterprise is under pressure to put the technology to business use and drive impact.

“I have the privilege of working hand-in-hand with AI leaders and data teams worldwide,” Jay Schuren, chief customer officer (CCO) at DataRobot, told VentureBeat. “What I’m hearing from them is that there is a tremendous amount of pressure and demand they are facing right now to get generative AI solutions out to the business. The major impression is that this technology is meaningful (not purely hype), and companies that move fast will have a real competitive advantage.”

However, the effort to move fast and deploy in production is encountering some roadblocks.

Solving generative AI problems

Companies across sectors have moved or are moving to complete their initial generative AI prototypes. However, the reality is that most of them are yet to realize tangible business value from these initiatives. 

Common challenges, as Schuren explained, include starting with the wrong problem, ecosystem lock-in, maintaining models and vector databases, not thinking through last-mile usage and not fully trusting the generative AI application’s outputs. 

“I’ve had CIOs tell me how there are new vector databases that have popped up and that they’re spelunking through logs to try to figure out who built them and what they’re supposed to be doing. I’ve also heard from CDOs that their teams have a roadmap of use cases but they can’t build them fast enough and get them out in a trusted way, and are playing in LLM playgrounds but don’t have the right way to get these solutions out in the market,” the CCO said.

To address these gaps, the company is now building upon its existing platform, enabling users to not only build generative AI solutions end-to-end (including solution development, backend hosting/monitoring and front-end hosting/monitoring) with a few lines of code and fewer personas, but also to integrate predictive AI models into these pipelines to audit generative AI outputs. 

For building and deploying gen AI solutions, the offering is providing a solution framework that allows users to integrate large language models, vector databases and prompting strategies of their choice with internal contextual (typically unstructured) data within DataRobot-hosted notebooks. 

This will give teams much-needed flexibility to use and compare different LLMs and other generative components to see what works best for their targeted use case.

Similarly, for building trust in the applications being developed, the offering will provide operational and data drift metrics as well as more specific generative AI metrics like toxicity and truthfulness to ensure applications stay “on-topic.”

“We are bringing the power of predictive sidecar models to validate and audit outputs of generative models. In addition, customers can define their own custom performance metrics that they want to use for monitoring things like truthfulness, topic drift and other use-case-specific metrics, as well as for tracking and monitoring LLM costs to ensure they don’t spiral out of control,” Schuren noted.

Finally, to streamline the feedback process and iteration on prototypes, DataRobot will host a Streamlit Application Sandbox. This will allow users to quickly prototype, build and deploy end-to-end applications/assistants to their business stakeholders.

Applied services bundled

When using these new capabilities, teams can also take advantage of DataRobot’s new enablement-focused applied AI services. This largely covers three areas: training to help leaders establish the level of generative AI proficiency needed to remain competitive; ideation and roadmapping to help teams go from use case ideation to implementation; and a trust and compliance framework to support responsible generative AI development and meeting existing and upcoming regulations.

“The new generative AI-focused services are offered both bundled and offered separately depending on the customers’ needs and the goals of the particular use case. Some teams want enablement throughout the process in the form of training, ideation/roadmapping, some want end-to-end delivery work, while others are looking for trust/compliance frameworks. We work with each customer closely to determine their needs and find the offer that they will be most successful with,” Schuren added.

Available right away

Teams looking to use the new generative AI capabilities and services from DataRobot can get started right away. The platform includes templated recipes with some best practices, which can later be customized to leverage the components that work best. Many organizations have started testing the offerings pre-launch, including Baptist Health South Florida and FordDirect.