Live shopping moves fast. Auctions end in minutes. Inventory sells out in seconds. A recommendation engine that’s even slightly too slow might as well not exist. Whatnot just acquired Shaped, a machine learning startup built specifically to solve that problem – and brought an entire Applied AI Research team along with it.
Financial terms weren’t disclosed, but the strategic logic is obvious. For an $11 billion marketplace processing over 500,000 hours of live video a week, getting recommendations right in real time isn’t a feature. It’s the product.
The Latency Problem That Drove the Acquisition
Traditional e-commerce recommendation engines have it easy. Catalogs are stable, products stick around, and customer preferences update gradually. Whatnot’s marketplace is nothing like that.
Unlike traditional e-commerce platforms, where product catalogs remain relatively stable, Whatnot’s marketplace is constantly evolving, and live auctions can end within minutes or last for hours. That makes standard recommendation infrastructure nearly useless. By the time a slow system identifies what a buyer might want, the show is over and the product is gone.
Six Years of Progress, Now Accelerating
Whatnot’s VP of Data and AI Emmanuel Fuentes said the company has spent six years reducing recommendation latency from about a day to just minutes, and expects the integration to push that closer to real time.
Getting from a day to minutes was hard. Getting from minutes to near-instant requires a fundamentally different approach – which is exactly what Shaped built. Shaped built a real-time AI personalization platform that delivers sub-50ms relevance across text, user, and session data, serving as ranking infrastructure.
That’s not an incremental improvement. It’s a different architecture entirely.
Who Shaped Is – and Why Murrell Matters
Founded in 2021 by Tullie Murrell, Daniel Camilleri, and Shai Bruhis, Shaped had raised a total of $10M in equity funding and is backed by Madrona, Y Combinator, Liquid 2 Ventures, Uncommon Capital, and Tribe Capital. Before launching Shaped, Murrell spent years building machine learning and recommendation systems at Meta – one of the most demanding recommendation environments on the planet, serving billions of users in real time.
A New Applied AI Research Group, Built From the Acquisition
As part of the acquisition, Shaped founder and CEO Tullie Murrell, along with nearly a dozen engineers and AI researchers, will join Whatnot. Murrell will lead the company’s newly formed Applied AI Research group.
That’s not just a talent acquisition attached to a technology deal – it’s Whatnot building a dedicated research function that didn’t previously exist, staffed by people who’ve already proven they can ship recommendation infrastructure at scale. Shaped’s technology was already live in production at QVC and Outdoorsy before Whatnot came calling.
A Platform Growing Faster Than Its Discovery Systems Can Keep Up
The urgency behind this acquisition becomes clearer when you look at how fast Whatnot itself is expanding.
Cross-category buying on Whatnot’s platform has grown 170% year-over-year, and the company has launched over 45 new product categories in just the first half of 2026. The company recently revealed that sellers have surpassed 1 billion orders, and last year Whatnot raised $225 million in Series F funding, giving the company a valuation of more than $11 billion after adding 20 million buyers over the past year.
That kind of growth compounds the discovery problem. More categories, more buyers, more sellers, more live sessions running simultaneously – and a recommendation engine that has to surface the right stream to the right buyer before the auction closes. Shaped’s technology allows for the analysis of these dynamic changes in real-time, with algorithms combining Large Language Models and machine learning to suggest the most relevant live streams to users based on their previous purchases.
Conclusion – Real-Time AI Is Now Table Stakes for Live Commerce
The gap between live shopping and traditional e-commerce isn’t just about video. It’s about time. Products expire. Buyer interest shifts mid-broadcast. A recommendation that arrives thirty seconds late is a missed sale.
Whatnot’s acquisition of Shaped is a direct bet that real-time AI personalization is what separates the platforms that scale from the ones that plateau. With Murrell’s team now inside the company and sub-50ms relevance infrastructure being integrated into a marketplace already processing half a million hours of live video per week, that bet looks well-considered.
For anyone tracking where live commerce AI is heading, this is the deal to watch.
Want to see how AI is reshaping personalization and recommendation across other consumer platforms? Read our breakdown of Equal AI’s $30M voice screening raise to see how real-time AI is changing the consumer experience beyond shopping.
