Sidd Motwani, Ian Anderson and Shivaditya Sinha dedicated years to creating the behavioral intelligence system that supports Spotify's recommendation engine. Known as Vector AI, the platform forecasts user intent and upcoming actions rather than depending only on prior activity. It drives nearly 90 percent of recommendations for the service's 800 million users.

The trio is now applying comparable technology to online retail through their startup, Malachyte. The company announced Thursday that it had secured $10 million in seed funding to expand its reach and recruit additional product and sales talent.

Malachyte originated from the observation that most digital stores treat every visitor uniformly. Personalization often relies on past purchases, broad demographic groups, or existing account data. As a result, new visitors encounter the same generic layout as others, while repeat customers receive suggestions based mainly on earlier buys instead of current needs.

The startup aims to address this gap by delivering real-time, intent-focused shopping experiences. Its platform employs what it describes as two-headed Vector AI to forecast the next desired item, capture overall preferences, and adjust continuously from live behavior.

The system begins to form insights before any click occurs by using available context at page load. Within one session it develops a clear view of both tastes and immediate goals, according to Motwani, Malachyte's CEO.

For instance, a search for heavy-duty boots followed by selections of steel-toed styles can elevate work pants and gloves while lowering dress shoes, all without requiring login details or history. Each further action refines the profile, making the experience increasingly relevant during the visit and on return trips.

Motwani notes that retailers already hold valuable customer data yet seldom apply it instantly. Every hover, click, scroll, search adjustment and cart addition serves as a signal, yet most platforms either ignore it in the moment or process it into segments later. The new system interprets these signals continuously so each step strengthens understanding of both preference and present intent.

Contextual cues are also underused, he added. A visitor on a phone late at night via an email link differs in mindset from the same person on a laptop in the morning, yet many systems handle them the same way.

The company has refined and tested the technology since 2024, collaborating with more than 20 enterprise clients in travel, grocery and retail before narrowing focus to e-commerce.

The platform first launched with Fun.com in fall 2025. It became generally available to Shopify merchants through native integration in June 2026, while larger retailers can connect via API.

Looking forward, Motwani sees the main opportunity in aligning merchandising and marketing around a shared view of customer behavior.

The seed round was co-led by Bessemer Venture Partners and Gradient, with participation from Harpoon Ventures.