Malachyte Raises $10M Seed to Bring Spotify's AI Recommendation Tech to E-commerce
Three ex-Spotify engineers just closed a $10 million seed round for Malachyte, a New York-based startup building behavior intelligence for e-commerce. The round was co-led by Bessemer Venture Partners and Gradient Ventures, with Harpoon Ventures also participating. The money will fund distribution scaling and senior commercial and product hires, according to CEO Siddharth Motwani.
The startup's pitch is straightforward: what if your shopping experience adapted to you the way Spotify's recommendations adapt to your listening habits? That's exactly what Malachyte's platform does. It uses what the company calls "two-headed Vector AI" to predict what product a shopper wants next, learn their general taste, and fine-tune continuously based on real-time behavior. No login. No cookies. No pre-existing customer data required.
"It starts forming before the first click, using the context available the moment the page loads," Motwani told TechCrunch. "Within a single session, we build a real read on both preferences and what someone is trying to accomplish right now."
A search for "heavy-duty boot" followed by two clicks on steel-toed boots moves work pants and gloves up the page and pushes dress shoes down. Every additional action sharpens the profile. The system reads every hover, click, scroll, search refinement, and add-to-cart event as a signal, updating the shopper's vector in real time.
The Spotify Connection: Building Vector AI at Scale
The three founders — Siddharth Motwani (CEO), Ian Anderson (CTO), and Shivaditya Sinha (COO) — spent years building the behavioral intelligence infrastructure behind Spotify's recommendation engine. Their system, called Vector AI, powers roughly 90% of Spotify's recommendations to its 800 million users across more than a billion items.
Anderson, a former Staff ML Engineer at Spotify, led a team of 100+ and helped redesign the platform's core recommendation systems. He also serves as Founding Chair of Spotify's AI Advisory Board, with multiple publications in user learning and recommendations. Motwani brings 12+ years of experience building data-driven products at Priceline and Spotify, where he took user data systems from zero to one and conducted additional validation work at Stanford. Sinha, the COO, comes from venture capital at a leading B2B firm in New York, with prior experience at WhiteHat Jr. and management consulting at BCG. He holds an MBA from Columbia Business School and London Business School.
The team started testing Malachyte's technology with more than 20 enterprise customers across travel, grocery, and retail in 2024. The platform went live in fall 2025 with Fun.com (operating as HalloweenCostumes.com) and has been generally available to Shopify merchants through a native integration since June 2026. Larger retailers can integrate through the API.
"We read it continuously, so each action makes the user's vector more confident about both preference and current intent," Motwani said.
How Two-Headed Vector AI Actually Works
Most e-commerce personalization systems work from historical purchases, demographic segmentation, or logged-in profiles. That means first-time visitors often see the same generic storefront as everyone else, while returning shoppers get recommendations based primarily on what they bought months ago rather than what they're looking for today.
Malachyte's approach is different. The two-headed Vector AI system understands preference and intent simultaneously — like an actual shopper does — by leveraging visuals and context rather than keywords or tags. One "head" tracks what the shopper seems to prefer broadly (their general taste). The other tracks what they're trying to accomplish right now (their immediate intent).
Contextual signals matter, too. A phone visitor arriving at 11 p.m. from an email link is in a different state of mind than the same person browsing on a laptop mid-morning. Most systems treat both identically. Malachyte doesn't.
The platform is persistent across channels. A signal picked up in search sharpens what that same shopper sees later on a product page, a category page, or even in an email or SMS the next day. The system holds up under pressure, too: sub-200ms latency under Cyber Monday-level load, auto-scaling with no reported outages, and a merchandiser control layer that gives operators the ability to set merchandising rules and overrides. Onboarding takes as little as seven days.
Real Results from Real Brands
Malachyte has been live with a growing roster of brands, and the early numbers are worth paying attention to:
HalloweenCostumes.com (Fun.com) saw a 31% lift in revenue per visitor after deploying the platform. Mark Bietz, Fun.com's Chief Marketing Officer, told Malachyte: "We haven't seen a merchandising platform provide this kind of growth, at scale, in years."
Brunt Workwear achieved an 80% lift in add-to-cart click-through rate. "We went from guessing what customers might want to showing them exactly what made sense in the moment," said Emilee Walch, VP of Digital Product at Brunt Workwear. "The ROI was clear quickly."
Jordan Craig, a fashion retailer, saw a 17% lift in new-visitor revenue per visit. That's significant given that 80% of their traffic comes from first-time visitors with no purchase history. "Our PDP was built to close the sale, but it wasn't built to grow it," said Robert Varon, VP of Digital at Jordan Craig. "Malachyte changed that."
Chad Greiter, Director of Digital Product at Simon, put it bluntly: "Malachyte is solving a problem most teams don't even know they have yet. Their approach to real-time user vector personalization, building a live profile for every visitor from the first click, without login or history, is exactly the kind of approach retail needs right now."
The Broader Context: AI Developer Tools Startups and India Investments
The funding landscape around AI has shifted dramatically in recent years. Customer acquisition costs have risen roughly 40% between 2023 and 2025, according to LoyaltyLion, and the average e-commerce brand now loses $29 per new customer after accounting for marketing spend and returns, according to SimplicityDX. Every unconverted visit compounds that loss.
"Brands are paying more than ever to acquire customers, then trying to convert them with a stack that only understands the fraction it already recognizes," Motwani said. "Our technology reads behavior instead of logins or cookies, so it understands every visitor before their first click, and keeps sharpening with every action."
This is why investors are paying attention. The broader trend around AI developer tools startups and India investments reflects a wider realization: the companies building infrastructure for intelligent commerce will determine who wins in the next decade. India, in particular, has become a major hub for AI investment, with deep technical talent pools and growing venture capital interest in AI-native startups.
McKinsey projects that by 2030, AI agents could orchestrate up to $1 trillion of U.S. retail sales. Counterintuitively, that makes understanding real shopper behavior even more valuable, not less.
"Consumers have so much less patience when website or search experiences are not tailored to their intent, and agents are increasingly shaping the buyer journey," said Maha Malik, Vice President at Bessemer Venture Partners. "Yet most commerce infrastructure is not built to understand buyer intent in real time, whether that's for a human or an agent."
What's Next for Malachyte
The seed round funds two priorities: scaling distribution of the existing infrastructure and hiring senior commercial and product leadership. The platform is already live with brands across travel, grocery, and retail, and the team has been iterating based on real-world feedback since 2024.
Motwani believes the bigger opportunity lies in bringing merchandising and marketing together around a single understanding of customer behavior. "Every unconverted visit compounds that loss," he said. "We're building infrastructure for how commerce actually works today — not how it used to work."
The question is whether the rest of e-commerce will catch up fast enough. The technology exists. The results speak for themselves. What remains to be seen is whether legacy personalization stacks can adapt before the window closes.
