A group of “LLM hipsters” are behind a new AI startup in Seattle emerging from stealth mode and aiming to rethink how retail companies engage with customers.
Spiffy incorporated a year ago and has raised about $6 million from investors including Point72 Ventures (which is growing its Seattle presence), the AI2 Incubator, Ascend, Sorensen Ventures, and J4 Ventures. It spun out of the AI2 Incubator at the Allen Institute for Artificial Intelligence (AI2) in Seattle.
Spiffy CEO Aniket Deosthali previously helped Walmart build its generative AI-powered shopping assistant. The company’s other co-founders include:
- Iz Beltagy, CSO, joined AI2 in 2017 and leads research on OLMo, an open generative language model created by AI2.
- Matthew Peters, chief architect, joined AI2 in 2016 and leads natural language processing and machine learning research projects. He was previously director of data science at Moz.
- Sameer Singh, CTO, joined AI2 in 2021 as an AI fellow, and is an associate professor at UC Irvine.
Spiffy’s core technology traces its roots to research at AI2.
“They had this vision that generic, large language models like ChatGPT just weren’t going to cut it when it came to solving substantive problems for businesses,” Deosthali said of his co-founders.
Spiffy is developing what it describes as “Outcome-Oriented Models,” a version of AI that is hyper-personalized and continuously improves toward a goal.
The idea is to equip retailers and their customer-facing workers with custom AI tools that don’t require extensive budgets to train models and run advanced graphics processing units.
Spiffy sees opportunities to help retailers of all sizes change the way they think about their customer service workforce — and not necessarily by replacing them, but by enhancing their workflow.
“Humans are often looked at as a cost center in organizations,” Deosthali said. “Our belief is that humans with AI can create opportunities to reimagine that fundamental assumption.”
Spiffy already has paying customers and the 4-person company is experimenting with different revenue models.
“Our fundamental approach has been to start with our customers’ dream outcome and work backwards from that,” Deosthali said.
Deosthali wouldn’t divulge much more about the company’s customers or product details.
“In a world with AI, every interaction with humans can be amazing, cheap, and revenue-generating,” he said. “That’s not possible today, but we believe that future is possible.”
Point72 Ventures partner Sri Chandrasekar said investing in Spiffy was a “no-brainer.”
“With the significant changes in the dynamics of consumer interactions with merchants and the way goods are purchased, we believe customer-facing companies need to utilize innovative technological solutions to remain relevant,” he told GeekWire in a statement. “The Spiffy team stands out to us due to their unique blend of industry expertise, inherent technical skill, and a humble understanding of the challenges involved in building a sizable company.”
