Azuryne
CPO and Cofounder
We set out to build Azuryne to solve the age-old problem of finding something to wear. Only the extremely wealthy can afford to hire their own personal stylist—we wanted to democratize fashion companionship.
I worked as Chief Product Officer, adopting a range of responsibilities that started with designing and developing our app and ended with overcoming communication barriers between LLMs and humans when we evolved our app into an LLM-driven SMS texting agent.
Azuryne - Digital Wardrobe App
We began as an innovative player in the digital wardrobe space. By utiilizing computer vision to extract clothes from photos, we introduced a novel and fast approach to logging clothes compared to the current wardrobe apps that required users to upload a photo of each individual garment in good lighting.
I designed the interface for logging clothes in Figma and built our app in React Native, our CEO developed the segmentation technology and our CTO worked on the backend in AWS.
By December 2023, we had a working app that was fast enough that the speed did not bother users, but we needed to pivot. In conducting usability tests with our current users, I learned that users were excited to use this app after all their clothes were logged, but expressed no interest in logging their entire wardrobe. Interestingly enough, these same users maintained that digitizing their clothes was fun and felt like a game but this sentiment wore off after the novelty disappeared. Our question we needed to answer was:
“How do we turn logging clothes into something that is fun every time?”
Azuryne - Instant Outfit Feedback
We decided we needed to create an app that brought our users value the second they used it. Users upload an outfit photo and the occasion they’re dressing for and Azuryne gives them a summary of what works, what to improve and additional information about their outfit like the type of aesthetic or occasions they could wear it for. While they receive actionable advice to improve their style, we passively grow their closet, addressing the problems with our previous product.
Usability tests for this new product revealed that users loved this product but had follow-up questions to the feedback they received. They needed a conversational aspect.
Azuryne - Your Fashion Bestie
We created an LLM-based chatbot connected to a phone number you could text to ask any questions about fashion to anytime and anywhere. In the background, we remembered these outfits and garments and information about user’s style preferences to create a stylist that knew your fashion inside and out. The product launched in April 2024 and amassed over 2,000 users in just 2 months. Users asked us anything from what to wear to school to help with purchasing new clothes and many referred to Azuryne as their “bestie”.
I was in charge with improving the conversations by reading messages, collecting and grouping bugs and friction in the conversation, and suggesting improvements on the product side. Additionally, I did prompt engineering on the inputs to enhance the chatbot’s performance to refer back to user’s previous styles and garments, create expressive style tags, and take on the right tone to keep the user’s trust and companionship. I also implemented the textline’s companion website, which a user could use their number to log into and view and edit all the information on the outfits and garments they sent. I designed and built this website in Figma and React.
