Whatnot Acquires ML Startup Shaped to Tackle Live Commerce's Hardest Recommendation Problem
Livestream shopping platform Whatnot has announced the acquisition of Shaped, a machine learning company specializing in real-time recommendation and search systems. The deal is designed to significantly strengthen Whatnot's personalization and product discovery capabilities as the platform scales across new categories and a rapidly growing buyer base. The acquisition was announced on Wednesday and marks one of the more technically focused AI investments in the live commerce space to date.
What Happened: The Core Facts
Whatnot, the livestream shopping marketplace launched in 2019, has purchased Shaped — a startup that built AI-powered recommendation infrastructure for businesses. As part of the deal, Shaped's founder and CEO Tullie Murrell, along with nearly a dozen engineers and AI researchers, will join Whatnot. Murrell, who previously worked at Meta before founding Shaped, will lead Whatnot's newly created Applied AI Research group.
Prior to the acquisition, Shaped's customer base included companies such as Outdoorsy and QVC. The startup's core technology combines existing customer data with large language models and machine learning to deliver highly personalized search and discovery experiences — a capability that maps directly onto the unique challenges of live commerce environments.
The Details: Why Real-Time Recommendations Are a Live Commerce Problem
Traditional e-commerce platforms operate with relatively stable product catalogs. Recommendation engines can be trained and updated on a daily or weekly cadence without significant impact on user experience. Live commerce, however, operates on an entirely different timescale — and that difference is at the heart of this acquisition.
As Emmanuel Fuentes, VP of Data and AI at Whatnot, explained to TechCrunch:
"Live commerce is a uniquely hard recommendation problem. Inventory changes by the second, shows start and end continuously, and buyer intent shifts throughout a show."
Fuentes noted that Whatnot has spent six years improving the speed of its recommendation engine, reducing recommendation latency from roughly a day to just minutes. The integration of Shaped's technology is expected to push those recommendations even closer to real time. According to the company, its systems already process more than 500,000 hours of live video and millions of real-time interactions every week, continuously feeding data back into its recommendation models.
Live auctions on Whatnot can end within minutes or run for hours, making the recommendation challenge fundamentally different from static product listings. Helping buyers discover relevant content while inventory, pricing, and demand shift dynamically requires an ML architecture purpose-built for low-latency, high-frequency data environments — precisely what Shaped was designed to deliver.
Einordnung: Why This Acquisition Matters for E-Commerce AI
The Shaped acquisition is not an isolated event. It takes place against a broader backdrop of significant growth at Whatnot and an industry-wide race to embed AI into resale and marketplace platforms. Competitors such as eBay and Poshmark are also actively integrating AI across their platforms.
Whatnot's scale makes the technical challenge — and this acquisition's strategic significance — easier to appreciate:
- The platform has surpassed 1 billion seller orders since its 2019 launch
- The company raised $225 million in Series F funding, achieving a valuation of more than $11 billion
- Over 20 million buyers were added over the past year
- More than 35 new product categories were launched last year, including art, golf, and vinyl
- More than 45 additional categories were added in the first half of 2026, with new subcategories continuing to roll out monthly
At that scale and growth velocity, a recommendation engine that operates on yesterday's data is simply not fit for purpose. The Shaped acquisition is an engineering and infrastructure investment as much as it is an AI strategy move.
For Shopware merchants and e-commerce operators watching from the sidelines, this deal highlights a direction of travel that extends well beyond livestream-native platforms: the expectation of AI-driven, real-time personalization is moving from premium feature to competitive baseline.
Practical Takeaways for Shop Operators and E-Commerce Managers
While Whatnot operates at a scale most Shopware merchants will not immediately match, the underlying principles of this acquisition translate directly to smaller e-commerce operations:
- Recommendation latency matters: The gap between a user action and a relevant recommendation response directly affects conversion. Even outside of live commerce, reducing that lag improves the shopping experience.
- Dynamic catalogs demand dynamic recommendations: Any shop running flash sales, time-limited offers, or frequent inventory changes faces a version of the same problem Whatnot is solving. Static recommendation rules quickly become stale.
- LLM-enhanced search is becoming table stakes: Shaped's approach of combining customer data with large language models for search and discovery reflects a broader shift. Semantic search and intent-based product discovery are increasingly expected by buyers across all platforms.
- AI team structure matters: The creation of a dedicated Applied AI Research group at Whatnot signals that AI is no longer just a vendor feature to be plugged in — it requires internal expertise and ownership to deliver competitive differentiation.
Outlook: What Follows from This Deal
With Murrell's team now embedded inside Whatnot's product and engineering organisation, the company is clearly positioning AI-driven recommendation as a long-term infrastructure investment rather than a one-time feature launch. The formation of the Applied AI Research group suggests ongoing development work, with Shaped's technology serving as a foundation rather than a finished solution.
For the broader e-commerce industry, the acquisition reinforces a clear signal: the platforms investing in real-time, ML-powered personalization at the infrastructure level are likely to widen their competitive advantage over those relying on conventional, batch-processed recommendation approaches. As live shopping formats continue to grow in relevance — and as buyer expectations for relevance and speed increase — the technical bar for recommendation quality will only rise.
Shopware operators and e-commerce developers should monitor how tools and plugins in their own ecosystem evolve to address these same underlying needs: faster, smarter, more context-aware product discovery at every stage of the buyer journey.