New Way Now: Stanford University powers safe AI exploration on campus

*Featured in this video:* Bhavana Tirukovalluri, Senior Director of Technical Services for University IT at Stanford University *Executive summary:* University IT, Stanford University’s central technology organization, set out to increase AI literacy on campus. Using Vertex AI, Model Garden, and Google Kubernetes Engine, University IT developed the Stanford AI Playground, a secure, user-friendly platform that empowers its entire community to explore the latest generative AI technologies. Built and launched in just two months, the AI Playground allows Stanford students, faculty, and staff to experiment with AI models while learning how to use them safely and for free. Across Stanford University, specific use cases are emerging fast, from operational and business support to professors applying it to analyze scholarly publications and summarize their research. The AI Playground has already seen immediate campus-wide adoption, serving more than 7,000 active users and processing 38 billion tokens — the tiny units of raw text and data gen AI models use — across millions of queries and interactions. *Challenge:* Stanford University believes in embracing new technologies while building in guardrails for responsible use. To increase literacy around AI on campus, University IT wanted to create a safe place for its community to explore and experiment with AI models and tools while balancing challenges associated with data privacy and security. *Solution:* University IT developed and launched the Stanford AI Playground in just two months — delivering a free and safe place where students, faculty, and staff can explore and experiment with AI. Powered by Vertex AI, the AI Playground provides easy access to a wide range of AI models and tools, with Google Kubernetes Engine delivering the necessary scale and flexibility to support the entire university community. In addition, Vertex AI Model Garden enables University IT to rapidly experiment with the latest AI models, test them internally, and make them available in the AI Playground within a week. Most crucially, Google’s commitments to AI privacy ensure Stanford maintains complete control — data is never shared externally or used to train models. *Results:* With the AI Playground, the Stanford community is already tapping into AI’s diverse potential. With more than 7,000 active users, the platform has already processed 38 billion tokens across millions of queries and interactions. By establishing this strong, secure foundation for AI, University IT is not only increasing confidence in AI across Stanford but also paving the way for more complex projects, including building AI agents. *Key takeaways and highlights from our interview with Bhavana Tirukovalluri, Senior Director of Technical Services for University IT at Stanford University:* → “Stanford AI playground is in a true sense a playground where the campus community can experiment and play with different LLM models and figuring out how does AI work. People are educated about hallucinations, prompts, how do you make it human-centered. There are more than 7,000 users that are using the AI playground actively — and there’s about 38 billion tokens that have been used so far.” → “It took us literally two months to bring up the AI playground because we had the framework with Google in place. It is accessible, it is geo-extensible based on the usage — the platform gives you that flexibility. We use Google Cloud for all our custom authentication platforms. That’s the underlying infrastructure. One of the key things for us was using Google Kubernetes Engine (GKE) to make it extensible and scalable.” → “It’s a feedback-driven application. Every week, we get requests, ‘Can you include this new model?’ It’s a constantly changing ecosystem. Having that flexibility of getting access to the LLMS through Model Garden, leveraging Vertex AI to help us build our agents as needed, and having a scalable infrastructure helped us grow our AI playground as the usage increased.” *Google Cloud products used: Vertex AI, Model Garden, Google Kubernetes Engine* *Learn more:* → https://www.stanford.edu/