StyleNova

A privacy-first AI wardrobe assistant that uses Amazon Nova 2 Lite to analyze clothing, personalize style advice, and recommend outfits through simple text chat.

Comment

Inspiration

StyleNova was inspired by a very common problem: many people already own enough clothes to dress well, but still struggle with daily outfit decisions. Questions like “What should I wear today?”, “Which colors actually suit me?”, or “What should I pack for a trip?” come up all the time, yet most people do not have access to a stylist or the time to manually organize their wardrobe.

I wanted to build something practical and personal: an AI wardrobe assistant that helps users make better use of what they already own. I also wanted the experience to feel trustworthy, so privacy became a core design goal from the beginning.

What it does

StyleNova is a privacy-first Android wardrobe assistant powered by Amazon Nova 2 Lite. It helps users:

  • analyze clothing photos and identify category, color, and style
  • analyze a user’s selfie to estimate skin tone and seasonal palette
  • generate outfit recommendations based on wardrobe inventory, weather, and profile data
  • answer text-based fashion questions like packing advice, event styling, or what to wear today

Instead of treating AI as a novelty, the app uses it to solve small but repeated everyday decisions in a simple and useful way.

How I built it

I built StyleNova as an Android app using Jetpack Compose with an MVVM + Clean Architecture approach. The app uses Room for local wardrobe storage, Retrofit for backend communication, and Hilt for dependency injection.

On the AWS side, I used:

  • AWS Lambda for the backend logic
  • Amazon API Gateway to expose REST endpoints
  • Amazon Bedrock to invoke Amazon Nova 2 Lite

The app sends structured requests to the backend for four main AI flows:

  1. Item analysis – a clothing image is sent as Base64 for moderation and classification.
  2. User palette analysis – a selfie is analyzed for skin tone and seasonal color palette.
  3. Outfit generation – local wardrobe data plus weather and user profile are used to create a recommendation.
  4. Text chat – users can ask for styling or packing advice and receive a conversational response with optional matching outfit suggestions.

A major architectural decision was to keep wardrobe images local whenever possible. Photos are stored on-device, and the backend uses them only in-memory during inference rather than saving them permanently to cloud storage.

Challenges I ran into

One of the biggest challenges was making AI output reliable enough for an actual app workflow. Free-form responses are not always ideal when the app needs structured data, so I designed prompts to return strict JSON for things like category detection, outfit selection, and chat suggestions.

Another challenge was AWS integration. Switching to Amazon Nova 2 Lite required updating the backend invocation flow and configuring the right IAM permissions for Bedrock inference profiles and model access. This was a strong reminder that building AI applications is not just about prompts—it also requires careful backend, deployment, and security setup.

I also had to balance helpfulness with privacy. Since wardrobe and selfie images are personal, I wanted the product to provide meaningful AI features without creating unnecessary cloud storage or data retention concerns.

What I learned

This project taught me that good AI products are often less about flashy features and more about thoughtful integration. I learned a lot about:

  • designing multimodal prompts for practical outcomes
  • structuring model responses for app reliability
  • connecting Android clients to serverless AWS backends
  • working with Amazon Bedrock and Amazon Nova 2 Lite
  • building privacy-aware user experiences for consumer AI

I also learned that a lightweight reasoning model can be enough for a meaningful real-world product when the experience is designed carefully.

Impact and future potential

StyleNova can help people make better use of the clothes they already own, reduce decision fatigue, and encourage more thoughtful wardrobe choices. That has practical value for students, professionals, and anyone who wants accessible styling support without the cost of a personal stylist.

In the future, I would like to expand the app with smarter wardrobe search, stronger personalization, and richer recommendation flows. But even in its current form, StyleNova shows how Amazon Nova can support a useful, privacy-conscious mobile experience built around real everyday needs.

Built With

Share this project:

Updates