AURA

• 

  Adding item

• 

  Aura's profile impression

• 

  Checking if item fits profile

• 

  Suggested accessories

Inspiration

AURA was inspired by my Lodge Captain, who struggled to understand her personal aesthetic. I also had friends who found it difficult to decide what to wear or pack when travelling. I wanted to help people understand their style through clothes they already love, without forcing them into one label.

What it does

AURA is an AI-powered personal style assistant on Telegram. Users upload clothing they like, and AURA builds a profile based on recurring colours, materials, shapes, patterns, and style traits. It can check new items, explain matches and contrasts, and generate personalised wardrobe inspiration.

How we built it

I built AURA with a Telegram interface, a FastAPI backend, Supabase for private data and image storage, and Railway for deployment. AI models analyse clothing images, create semantic representations, and generate reference images. Application logic validates the results and produces evidence-based profiles and explanations.

Challenges we ran into

Early feedback could sound vague or repetitive. User testing also revealed issues with uploading items individually, navigating messages, and correcting AI analysis. I improved the experience with album support, clearer explanations, wardrobe browsing, guided corrections, and better suggestion variety. I also resolved AI provider and deployment configuration issues.

Accomplishments that we're proud of

I am proud that AURA became more than a simple AI prompt. It combines AI analysis with structured profiles, explainable matching, private storage, user corrections, and safety controls. Feedback from test users directly shaped the final experience.

What we learned

As a Year 3 Software Engineering student at SMU with little previous AI experience, I learned how to integrate AI into a complete application. I also reinforced my knowledge of software architecture, testing, security, databases, deployment, and user-focused design.

What's next for AURA

Future improvements could include virtual clothing previews, real product discovery, stronger personalisation from consented feedback, broader evaluation, and production monitoring. AURA will continue focusing on helping users understand their taste while keeping them in control.

Built With

• computer-vision
• fastapi
• image-generation
• postgresql
• python
• railway
• supabase
• telegram
• vector-embeddings