Ex-Spotify employees raise $10M to bring the AI behind its recommendations to e-commerce

Three people who spent years building the guts of Spotify's recommendation engine — the thing that quietly decides what 800 million users hear next — just pointed that expertise at online shopping instead. Sidd Motwani, Ian Anderson and Shivaditya Sinha helped create Vector AI, the system behind roughly 90% of Spotify's suggestions. Their new startup, Malachyte, announced a $10 million seed round on Thursday, co-led by Bessemer Venture Partners and Gradient, with Harpoon Ventures also chipping in.

The pitch is simple, even if the tech underneath isn't: most e-commerce personalization is stuck looking backward. A first-time visitor gets the same generic homepage as everyone else. A returning shopper gets recommendations based on what they bought six months ago, not what they're clearly hunting for today. Motwani's argument is that retailers already sit on a goldmine of real-time behavioral signals — hovers, scrolls, search tweaks, add-to-carts — and mostly ignore them until some overnight batch job lumps everyone into stale segments.

Malachyte's answer is what it calls a "two-headed Vector AI," which starts forming a read on a shopper before they've even clicked anything, just from page-load context. Motwani gave a concrete example: search "heavy-duty boot," click two steel-toed pairs, and within that single session the system starts pushing work pants and gloves higher while dress shoes drop down — no account, no login, no purchase history required. The model keeps sharpening with every action, and apparently carries some of that understanding into a person's next visit too. He also flags context most platforms flatten out, like the difference between someone browsing on a phone at 11 p.m. from an email link versus the same person on a laptop mid-morning.

This isn't a lab experiment dressed up for a funding announcement. Malachyte has been testing since 2024 and ran pilots with more than 20 enterprise customers spanning travel, grocery and retail before narrowing its focus to e-commerce specifically. Fun.com was its first live deployment in fall 2025, and since June 2026 the platform has been generally available to Shopify merchants via native integration, with an API for bigger retailers who need something more custom.

The fresh capital is earmarked for scaling distribution and hiring product and commercial leadership — the unglamorous but necessary work of turning a promising demo into something enterprise retail actually trusts at scale. Motwani's longer-term bet is that merchandising and marketing teams, who currently operate on separate data and separate assumptions about the customer, should eventually be working off the exact same real-time behavioral model.