AI Cloud Innovation Center optimizes speed, reduces cost for ASU’s WaterBot

For the past 19 years, the Colorado River system, which supplies 36% of Arizona’s total water use, has experienced extensive drought conditions. While more snowfall and rain this winter improved the river’s health, the U.S. government has called for Arizona to take an 18% cut from its total Colorado River water allocation for 2024.

With limited water resources, it's important for Arizonans to stay informed about water and why conserving it is essential. Enter WaterBot, a chatbot to deliver facts on water management in Arizona, including drought conditions, water conservation and more.

Developed by Arizona State University (ASU) Associate Professor Stephen Carradini, WaterBot was created with support provided by ASU’s Arizona Water Innovation Initiative and the Impact Water - Arizona program, courtesy of the State of Arizona and the Virginia G. Piper Charitable Trust. Arizona Water Innovation Initiative is a statewide program led by the Julie Ann Wrigley Global Futures Laboratory in collaboration with units across the university, such as the College of Integrative Sciences and Arts (where Carradini is housed) and the Ira A. Fulton Schools of Engineering.

While Carradini and research assistant Briana Rajan built WaterBot, set up the structures, and performed checks from August to December 2023, they wanted to find ways to optimize the speed of the chatbot while being efficient with funding. That’s when Carradini teamed up with the ASU Artificial Intelligence Cloud Innovation Center, powered by AWS (AI CIC) to harness the power of AWS’ suite of products to assist with the deployment of WaterBot. This partnership ultimately resulted in minimizing costs and catalyzing the sharing of knowledge for Arizona residents.

Overcoming technical hurdles

In late 2023, Carradini was looking for ways to optimize the bot's performance and better manage operational costs for the chatbot. Carradini found the answer to his search at the AI CIC.

“We really wanted to make sure we’re using the best practices that we can, giving the best product we can put forward in the interest of the state of Arizona and the citizens of Arizona,” Carradini said. “We wanted to optimize speed and minimize costs—which you usually can’t do—but we found out that we could do both with AWS. And so, we jumped on that,” Carradini said.

The AI CIC team, including Senior Solutions Architect Arun Arunachalam and the AI CIC’s student workers, provided support to Carradini when it came to WaterBot’s deployment. They introduced him to AWS products to bring flexibility and affordability to the project, including a serverless, pay-as-you-go compute engine called AWS Fargate and Amazon Bedrock for generative AI services and development.

“One of the big things in cloud computing is ‘pay as you go’ and ‘scale on demand,’” Arunachalam explained. “So, if you have 500,000 users coming to your website, you pay for 500,000. If you have one, you pay for one.”

This shift was pivotal, enabling the team to cut costs while maintaining performance. By packaging the application into containers and deploying them on AWS, the WaterBot's base running costs dropped dramatically—from $112 per month to around $35 on AWS—in less than three weeks. “We immediately saw a large percentage decrease, and now that we’ve continued working with [AWS], we’re seeing consistent month-over-month great improvement over initial running costs,” said Carradini.

The project was able to move so quickly because Carradini had the chatbot built on a firm technical foundation. “We did not have to start this project from scratch, like doing initial analysis, development and testing – all that was already done by the professor and his team,” Arunachalam said. “It was more on improving the efficiency in the deployment, focusing on his needs.”

The bot is now effectively serving the Arizona public and beyond, serving as a go-to resource for information on water management in Arizona, including drought conditions, water conservation and more.

The power of collaboration

Looking ahead, Carradini and Arunachalam remain optimistic about how their collaborative partnership can continue to flow. “We are hopeful that there will be other options in making bots, like solar bot or heat bot, or things like this from our lab,” said Carradini.

The experience gained from this project will undoubtedly influence future endeavors, both in terms of technical implementation and interdisciplinary collaboration. Some of the techniques and practices that were in WaterBot’s code are already being used on other projects. Carradini hopes that others will build their own chatbots using their secure, open-source code.

As the WaterBot continues to evolve, it will serve as a model for future projects aimed at leveraging technology for the greater good of the ASU community, the state of Arizona and beyond.