Using AI for Accurate Information

Kartikeya Vats (MSAI '23) wants you to get what you pay for.  

Vats is a senior AI and data scientist at Target. He leans on lessons from Northwestern Engineering's Master of Science in Artificial Intelligence (MSAI) program to build predictive models that identify if items are out of stock on the sales floor despite being listed as available in the store's ledger system. 

His mission is to create happy, loyal shoppers while simultaneously strengthening the company’s supply chain.  

“If you believe an item is present on the shelf of the store but it’s not, this misinformation creates confusion,” said Vats, who has been with Target full-time since January. “It won’t trigger replenishment from our distribution centers or from the back room. That impacts the customer experience.”  

A variety of factors – ranging from technology glitches to shoplifting – cause this type of inventory problem. Vats builds AI models that weigh these issues and trigger resupply efforts at the right time.  

Seeing the immediate impact of his work is rewarding, he said.  

“Everything gets measured, and the feedback loop for that impact measurement is pretty quick,” said Vats, who interned at Target as a MSAI student. “What I learned during my time in Northwestern is to constantly generate better ways of solving problems. I have been able to contribute more to problem solving than I would have been able to otherwise.”  

Vats was six years into his data science career, working at IBM, when he decided to further his education. The Northwestern brand name was a strong draw, he said. The MSAI program’s small cohort and advanced curriculum also played significant roles. 

“The whole setup is designed well for you to learn,” Vats said. “Northwestern had a lot of focus on modern AI techniques, and there was a lot of peer learning.”  

That peer learning proved invaluable, Vats said. His cohort ranged from some relatively new to AI to those with more than a decade of professional experience.  

Vats said he benefited from learning how others see business problems and apply AI as a solution.  

“I improved myself by drawing from other people's experiences,” he said. “It helped me understand a whole new spectrum of things that were happening around me in the field of AI.”  

Included on that spectrum is a better understanding of the dangers of AI.  

Some may worry AI will replace humans in the workforce, but Vats' concern revolves around people not understanding how AI works. AI wasn't designed to be factually perfect, Vats said. The goal, in his eyes, was for it to be coherent and creative. 

That creativity can be valuable in many ways, but it also presents challenges.  

When a large-language model (LLM) like ChatGPT doesn't have a full answer to a query, it sometimes fills in the gap by inserting what it considers a logical conclusion based on its training. The problem is sometimes, that gap is filled with wildly inaccurate information.  

“The fact is that, while impressive, they should not be completely relied on as a source of information,” Vats said. “They still hallucinate a lot.”  

Vats developed the knowledge to understand LLM's — both their potential and limitations — in MSAI. Classroom lessons were key to his growth, but they were just one part of his MSAI experience. That's the message he shares whenever he interacts with the program's prospective students.  

“Look beyond just doing well in classes and getting good grades,” he said. “Network more. Nothing will be handed to you, but the opportunities are there. Be mindful and grab them.”