Julie Bornstein on Daydream: Reinventing Fashion Discovery with AI

Daydream's Vision for Fashion Search

• 💡 Daydream is an AI-powered fashion search engine designed to replicate the in-store conversational shopping experience online.
• 💬 It leverages Large Language Models (LLMs) to enable natural language interaction, allowing users to describe what they're looking for in detail, like a "revenge dress for a wedding in Paris in the vibe of Saltburn."
• ✅ The platform helps users find, save, and share items, ultimately directing them to purchase on the brand or retailer's site.

Journey Through E-commerce Innovation

• 🚀 Julie Bornstein's career, spanning Nordstrom, Sephora, Stitch Fix, and The Yes, has consistently focused on using emerging technology to simplify online shopping.
• 🧠 Early efforts at Nordstrom involved basic keyword search for e-commerce, while Sephora explored mobile, social, and a conversational platform called "Beauty Talk."
• 📊 Stitch Fix utilized data science, machine learning, and computer vision for personalized recommendations, though limited by its owned inventory model.
• 🎯 The Yes was built as a fashion recommendation engine for engaged customers, later acquired by Pinterest, further refining the approach to personalized discovery.

Daydream's AI-Powered Shopping Experience

• 🖼️ Daydream offers multimodal input, allowing users to start searches with text queries, photo uploads, or voice commands, which often leads to more conversational requests.
• ✨ It provides an iterative shopping experience, enabling users to refine searches based on images and text, such as finding a similar item but in a different color or sleeve length.
• 🔑 Each user receives a Style Passport, an explicit and editable repository of their preferences, which helps the system predict desired items and personalize search results.
• 🛍️ The platform consolidates products from all fashion brands and retailers in one place, offering a comprehensive view of new items without requiring users to visit multiple sites.

Technical Approach and Market Position

• 🔬 Daydream moved away from directly using large LLMs due to inconsistency and latency, instead building smaller, more responsive "mini models" tailored to fashion dimensions.
• ⚙️ These mini models are trained using information from larger LLMs and augmented with AI-derived insights about product catalogs (e.g., occasion, body type, weather suitability).
• 🛡️ While large model providers like OpenAI and Google are entering shopping, Daydream's defensibility lies in its deep domain understanding of fashion and its ability to provide nuanced, taste-based recommendations.

The Future of Shopping with AI Agents

• 🤖 The web is shifting towards agent-driven discovery, where AI agents perform tasks for consumers, fundamentally changing how brands drive attention and sales.
• 🗣️ Daydream envisions a future with a mobile-first, voice-first interface where a fashion agent knows user preferences and upcoming events, suggesting relevant items without overwhelming choices.
• 💡 The goal is to make shopping decisions more confident and efficient, reducing returns and facilitating the discovery of new brands through a simple, powerful interface.

Lessons from a Serial Founder

• 📈 Julie Bornstein finds her second venture less stressful and vulnerable than her first, with fundraising being a smoother process due to prior experience.
• 🤝 A key lesson was the importance of holding out for the right technical talent, especially a CTO who embraces experimentation in the rapidly evolving AI landscape.
• 🌱 Daydream operates with a growth mindset, actively exploring new tools and techniques to leverage AI for increased efficiency and faster development, enabling a smaller, more agile team.