Roadmap to a Holistic AI Strategy: The Dataiku-Snowflake Partnership

Dataiku Product Marie Merveilleux du Vignaux

Attendees to Thomas Scholz’s talk during the Frankfurt Roadshow of the Dataiku 2023 Everyday AI Conferences were promised a product demo worth potentially millions in ROI.

Scholz, Director of Sales Engineering at Snowflake, led a compelling conversation alongside Sofiane Fessi, Regional Vice President of Sales Engineering at Dataiku, taking a closer look at the critical role of Generative AI in the business landscape, how the synergistic relationship between Dataiku and Snowflake systems forms the foundation of a strong long-term data strategy, and how companies and industries can design similar holistic strategies.

EAI frankfurt snowflake session

Snowflake’s founders realized that, at the time of the company’s inception, no data management platform really leveraged the benefits of public clouds for data management. “Things like elastic scalability, unlimited scalability, pay for what you use, the kinds of metrics that come to mind — they hadn’t been available with the traditional data management solutions on the market,” Scholz says. By developing a brand new architecture that separated the storage layer from the compute layer, the benefits for data managers increased dramatically. On the holistic nature of data management going forward, Scholz says, “It’s about artificial intelligence, Generative AI, and how companies can use these new technologies to bring their own data and analytics strategy to the next level.”

Design Effective Data and Analytics Strategy With the Power of AI

Snowflake was designed with the basic understanding that traditional data management use cases like data warehousing and data lakes aren’t the limit, and today’s industries want to increasingly collaborate with data in mind. Industry leaders are starting to understand that data no longer lives in closed ecosystems. To remain competitive, the data needs to not only be shared between team members or internal departments, but also between larger partners in a wider ecosystem or different companies that share a longer supply chain.

AI is really crucial for enterprises to be successful, to stay competitive. If they don’t think about AI strategies, their competition will. Over time those companies that do it well will gain a competitive advantage.

The excitement and energy surrounding the further integration of AI into business processes isn’t lost on Scholz. “The speed of innovation here is immense,” he says. Generative AI is starting to evolve from making decisions based on a given dataset to finding deep relationships, correlations, and connections with the data that aren’t immediately apparent.

Snowflake, an industry leader in cloud-based data storage and analytics and major Dataiku partner, recognizes that in addition to using AI as a conversation copilot, the productivity upside is massive. If companies already have strong data management solutions in place, Generative AI can make analysts even more productive by helping them focus their attention on certain parts of the data instead of poring over massive amounts of information manually to perform analysis or gain meaningful insights.

The natural evolution of integrating Generative AI and large data sources is that over time, the models themselves improve and produce insights that are increasingly relevant and tailored to your organization. Scholz says, “AI can only work for you if you combine the AI models with the data that’s in your own organization.” It’s only after an organization brings the LLM to the data, enriches the model with data, and does some feature engineering, that there can be real value to the organization.

Stay Data Privacy Compliant and Enrich Data Strategy Simultaneously

The foundational concept of data governance, routing, and access forms the backbone of the LLM Mesh. Within Snowflake, organizations can build proprietary models and deploy them in a container in a secure environment, in effect bringing the model to the data, instead of the other way around. Bringing the model to the data keeps the datasets safe in the organization and within a protected infrastructure.

Keeping datasets safe and regulatorily compliant is of the utmost importance for businesses in today’s landscape, and utilizing AI in the Dataiku LLM Mesh is a complementary tool to accomplish these goals.

Fessi of Dataiku took an even deeper dive, again alluding to the product demo that could be worth millions, in the context of a financial institution. “If we take a bank and a database of all of the customer transcripts, we know how vital and paramount customer satisfaction is. What we want to do here is get information from every single call. What was the call about? Was the issue resolved? What was the sentiment of the call?”

people working on computers

The value-add solution of combining raw datasets that exist in Snowflake and a robust AI solution from Dataiku (and the million-dollar demo) is how the insights are collected, automated, and continuously improved. By aggregating the responses to the above questions into integrated dashboards, banks can quickly generate insights and more readily share them across the organization. Additionally, because of the complementary nature of the LLM mesh, the LLM’s performance and relevance can be monitored over time.

The LLM helps the customer service to improve, and the customer service helps the LLM to improve. Human-machine symbiosis. How beautiful is that?

Governing personally-identifiable information (PII) is also a critical component in a data management system. “That’s when the power of Dataiku and Snowflake in the LLM Mesh comes into play,” Fessi explains. The raw data containing PII is stored in Snowflake and flagged by Dataiku as GDPR-sensitive information. Afterwards, the PII is removed using Presidio data protection software from Microsoft. The now PII-free data is pushed into another Snowflake database for customer service agents to review. From here, data scientists can use the Dataiku Prompt Studio to design scenarios to send Slack summaries to customer service agents to review the AI scoring, starting the feedback loop and continuously improving the LLM — all while remaining PII-compliant throughout. Want to go further? Check out other Dataiku solutions designed specifically for financial institutions.

Dataiku Snowflake LLM architecture

Key partnerships like the one between Snowflake and Dataiku allow for the creation of powerful LLM solutions, forming the foundation of a strong data and AI strategy. You’ll be able to unleash the power of Generative AI, keeping complete control of your data and protecting PII. You’ll be able to quickly share deep insights across your entire organization, and ensure that the models get better and better over time.

 

You May Also Like

Foster Trust in Your GenAI Deployments

Read More

Dataiku: The AI Platform of Choice for CIOs and IT Leaders

Read More

Take Your Data Strategy to the Next Level With FAIR Analytics

Read More

Taking PRIDE in Responsible AI via Data Collection & Analysis

Read More