AI for Personalized Brand Engagement

One size does not fit all. This simple concept was the motivation behind a startup called Spiffy, which leverages AI to create a hyper-personalized retail experience.

“We started as a bunch of AI researchers working on different aspects of large language models and natural language processing, but one thing that was frustrating was seeing people use one giant LLM to do everything for everyone,” says Spiffy co-founder and CTO Sameer Singh, also a professor of computer science in UC Irvine’s Donald Bren School of Information and Computer Science (ICS). “We knew from our experience that if you could specialize these language models for specific use cases, you could better control their behavior.”

As an example, Singh talks about how everyone uses ChatGPT, which is the same tool regardless of the user or query. “Lots of brands care about their voice. There’s a certain personality they want to exude, and they want to be able to control, from a legal or compliance point of view, what the agent is saying,” he explains. “But if you just use one single model across all merchants and across all agents, you’re not quite able to guarantee these things.” Spiffy aims to address this.

Hyper-Personalized AI

Singh, an authority on safe and reliable natural language processing, teamed up with a trio of researchers to launch the startup. Aniket Deosthali, the CEO, comes from a retail and e-commerce background, while chief scientist Iz Beltagy and chief architect Matthew Peters come from the Allen Institute of AI. Beltagy is a leader in AI training and Peters is a machine learning pioneer who invented the first foundation model.

“The idea was to use our academic expertise to create e-commerce products where AI is at the heart of it,” says Singh. By working to understand both the merchant and user context, they’re creating unique experiences in which the merchant still maintains control of the brand voice. “We believe in creating very customized AI that is specific to each brand and each merchant.”

The idea was to use our academic expertise to create e-commerce products where AI is at the heart of it.  — Sameer Singh

Spiffy incorporated toward the end of 2022 and raised about $6 million from investors in its first year. It now has a fully functioning product and customers. Also, in addition to the four founders, more than a dozen people now work at Spiffy, including UCI alum Prasad Mahendra, who earned his bachelor’s degree in computer science in 2001.

“I’ve been fortunate in my career to be a founding member of several successful startups since graduating from UCI, and I’m incredibly excited about Spiffy,” says Mahendra. “We have a remarkably strong team and are working on something truly groundbreaking in the field of generative AI.”

Leveraging AI Expertise

Spiffy is leveraging its team’s deep-rooted computing expertise in setting itself apart. “We’re a deeply technical company, and we have expertise in how to train and deploy these large language models, which really differentiates us from other people who are building these products,” says Singh.

He adds that having academic researchers is another plus. “People in academia really care, especially in my area of safe and reliable AI. We’ve been thinking about how to evaluate LLMs properly to ensure they’re robust and safe, and those ideas carry over into the company.”

So, while the long-term goal is to fundamentally revise the retail experience, security and reliability remain at the foundation of this work.

“We are at a growth stage right now,” says Singh, “so we are getting more and more customers and continue looking for people who can help us deploy AI to many users in a safe way.”

— Shani Murray