
ASU students create data science AI tutors
Why are students less interested in learning? The problem statement was presented by team M^6 at the AI Accelerated Spark Challenge, sponsored by NVIDIA, TiE Phoenix and AZNext on Monday, June 23.
Their AI tutoring solution earned them first place as the only all-women team competing. Team members included graduate students Darlene La Mere, Neha Tiwari, Pooja Pal, Rachel Garcia, and Sai Sudha Piratla.
“Things have changed a lot in the past two, three years with LLMs,” said Tiwari. So, at this point in time, being able to create something like an LLM, which is using RAG, using NVIDIA, using the SOL supercomputer, all of this was a tremendous learning opportunity for us and made us feel confident.”
While they tackled their problem statement on learning, they found that they learned a lot themselves, and that the experience better prepared them for their various career projections.
The AI Accelerated Spark Challenge was a week-long virtual hackathon held from June 23 - 30 over Zoom. The hackathon was led by ASU’s AZNext with co-sponsorship from TiE Phoenix. The hackathon concept relied on NVIDIA's education materials and support for GPU Acceleration. Hosted by Enterprise Technology, the challenge saw over 75 students participating from around the world as ASU students logged in from home for the summer. Many participants were first time hackers or were paired with students for the first time; all three winning teams were newly paired students, a testament to their ability to work together and problem solve in an accelerated context.
In partnership with NVIDIA, students were invited to create their own AI tutoring system using machine learning models, large language models (LLMs), and other data science tools. Their tutoring system would primarily focus on guiding their user how to GPU-accelerate data science.
Second place winner was team Gigabits, made up of students Ajay Yadav, Amogh Shetty, Debopam Banerjee, Sagar Sinha, Susrik Mukherjee. Finally, the third place winner was team R1, made up of students Amrit Singh Johal (AJ), Aryan Yeole, Aashir Javed, Bichen Pang, Vihaan Patel and Yohannes Ibsa.
Students had the option of either using MyAI Builder or their own custom interface to build a tutoring solution focused on GPU-accelerated data science. At a minimum, every team was competing to create the best and most engaging AI tutoring experience.
The projects, primarily coded with Python and built on ASU’s Sol supercomputer, gave students the unique opportunity to experiment with scaling machine learning workflows.
Rachel Garcia, winning team member of M^6, said her experience competing helped give her confidence as she goes into her programming field, which is constantly changing with AI innovations.
“I think doing this AI challenge has helped because I was able to collaborate with so many people that have so many different skills, and they’re so great at what they do,” Garcia said.
Students as far as India and Vietnam logged in the Zoom session to create 16 teams all working to create the most efficient AI tutoring system using their available resources.
The Spark Challenge was organized under the leadership of Raghu Santanam, Sr. Associate Dean at the W. P. Carey School of Business and Executive Director of AZNext, Rob Buelow, program director for AZNext, Jeania Kimbrough, research project manager with AZ Next, Gil Speyer, director for computational research accelerator, and Olivia Herneddo, lead experience designer with Enterprise Technology.
The five-day challenge featured two days dedicated to “bootcamp,” in which students learned more about AI on the Sol Supercomputer, and were able to workshop MyAI Builder along with various NVIDIA tools. On day three, they began coding. Mentors from ASU included Gil Speyer, Juan Jose Garcia Mesa, William Dizon, and Mansi Patel. Mentors from NVIDIA included solutions architects Amanda Butler and Zoe Ryan.
Gigabits team member Debopam Banerjee said the challenge helped him better understand what he wants to focus on in his career field.
“Working on this project made me realize I wanted to work on AI agents that actually help people, and not just automate things,” said Banerjee. “I’m more interested in how UI or UX shapes how people interact with these AIs long term.”
Students were introduced to various NVIDIA LLMs along with several open source libraries accessible through Sol. The libraries included python codes CuPy, Numba, RAPIDS, and more.
Gil Speyer looked forward to the innovations surrounding personalized AI learning with all the resources available to the students.
“We’re trying to meet students where they are,” said Speyer. “That’s why we’ve provided this full menu for them to choose from. You can make a whole ecosystem out of it for any student to learn these types of approaches.”
The top three winning teams earned scholarships ranging from $1000 - $3,000, and all competitors earned digital credentials.
Judges for the final project pitches included John Almasan, global head of AI and emerging technology with TIAA, Madhu Sudhan Reddy Gudur, chief AI officer at Axyo Inc., Steve Niemi, NVIDIA senior account manager, Raghu Santanam, Sr. Associate Dean at W.P. Carey, Gil Speyer, director for computational research accelerator, Rohit Taneja, senior product consultant at PetSmart, and Louise Tung, clinical assistant professor at W.P. Carey. Together, these evaluators bring a powerhouse of expertise across AI innovation, product strategy, academia, and industry leadership—making them uniquely equipped to assess and award student collaboration.
While many computer science, data analysis, and various technology-based fields are repeatedly shifting with new and improved AI software, all three winning teams of last week’s Spark Challenge said the competition really helped prepare them to take on AI in their respective fields.
“There is a lot of scope for new things to be discovered and created, and that’s what we got an opportunity to do,” Tiwari said.