Whatnot acquires Shaped to speed up live-shopping recommendations

Whatnot acquired Shaped, a machine-learning company specializing in real-time recommendations and search, to accelerate live-shopping discovery. Shaped founder Tullie Murrell and nearly a dozen engineers will join Whatnot to lead its Applied AI Research group. The deal aims to reduce recommendation latency from minutes to real time, addressing the challenge of ranking constantly changing live auction inventory.

Whatnot https://www.whatnot.com/?ref=runtimewire has acquired Shaped https://www.shaped.ai/?ref=runtimewire , a machine-learning company for real-time recommendations and search, in a deal aimed at moving live-shopping discovery closer to the speed of a live auction. TechCrunch reported https://techcrunch.com/2026/07/15/whatnot-acquires-shaped-to-power-real-time-live-shopping-recommendations/?ref=runtimewire the acquisition on July 15, and Whatnot said in its announcement https://blog.teamwhatnot.com/unitedstates/whatnot-acquires-shaped-ai?ref=runtimewire that Shaped founder Tullie Murrell and nearly a dozen engineers and AI researchers will join the company. Murrell will form and lead Whatnot's Applied AI Research group. Whatnot did not disclose deal terms or whether Shaped's standalone product and customer contracts will continue. TechCrunch identifies Murrell as Shaped's founder and CEO and says he worked at Meta before launching the company. Shaped, a Y Combinator https://www.ycombinator.com/companies/shaped?ref=runtimewire -backed startup, built infrastructure for real-time personalization and search. That fits Whatnot's hardest product problem: live supply changes constantly. A seller can go live, run an auction, sell out inventory, switch categories, or pull buyers into a fast-moving bid cycle in minutes. Traditional marketplace search assumes the catalog is relatively stable; Whatnot has to rank shows, sellers, categories, and listings while the thing being ranked may disappear as a show ends. A relevance engine for live discovery Whatnot frames the deal as infrastructure for discovery and personalization. The company says its systems already process more than 500,000 hours of live video and millions of real-time interactions each week. TechCrunch adds that Whatnot has spent six years cutting recommendation latency from roughly a day to minutes, and expects Shaped to push those recommendations closer to real time. TechCrunch reports that Shaped's customers included Outdoorsy and QVC. The announcement does not say what happens to those customers after the deal. Scaling past collectibles makes ranking harder Whatnot says sellers on the marketplace have passed 1 billion orders. It began in 2019 as a Funko Pops marketplace, now spans hundreds of categories, sells more than 20 items per second, and says fashion is its largest category by order volume. Source https://blog.teamwhatnot.com/unitedstates/1-billion-orders?ref=runtimewire For financing context, Crunchbase News https://news.crunchbase.com/venture/ecommerce-unicorn-whatnot-raises-seriesf/?ref=runtimewire reported that Whatnot raised a $225 million Series F at an $11.5 billion valuation, bringing total funding to about $968 million since its 2019 inception. Murrell's new Applied AI Research group is built for that constraint. Shaped spent its independent life packaging ranking and retrieval so product teams did not have to stitch together vector databases, feature stores, rerankers, and custom pipelines. Whatnot now wants that capability inside its marketplace, where every extra minute of ranking delay can send a buyer to the wrong show, or to no show at all.