Luvera

Reimagining skincare, one ingredient at a time

Comment

Inspiration

Skincare is often expensive, making it inaccessible to college students and people on a budget. Furthermore, people struggle to find products that work for the specific skin type and concerns. While many skincare apps and resources exist, they often prioritize brand marketing over ingredient efficacy.

We wanted to build a solution that:

  1. Focuses on budget-friendly, but effective products
  2. Uses AI to personalize skincare recommendations
  3. Help users navigate ingredient lists with transparency
  4. Remove the guesswork from building a skincare routine

What it does

Luvera simplifies skincare selection and routine building by:

  1. User Consultation: The user uploads 3 photos of their face (front, left, and right)
  2. AI Skin Analysis: Our AI determines the user's skin type and detects problem areas (e.g. acne, redness, hyperpigmentation, etc.)
  3. Ingredient Matching: Based on the skin issues detected, our AI determines the best skincare ingredients known to treat those concerns
  4. Product Selection: Our algorithm cross-references ingredients with a pre-approved database of affordable, effective products that we scraped from EWG.org and INCIDecoder to find the best options within the user's budget while also taking into account ethical health standards (cruelty-free, non-toxic)
  5. Routine Generation: The AI structures a personalized routine based on the recommended products
  6. "Looksmaxx" rating: We added a humorous "Looksmaxx" feature for a lighthearted analysis of facial attractiveness

How we built it

Frontend: NextJS
Styling: Tailwind CSS, Tailblocks
UI Enhancements: Framer, HeroUI, MagicUI

Web Scraping: Python, Playwright
AI Skin Analysis: Hugging Face (DeepSeek)

Product Database: EWG.org, INCIDecoder (trusted sources for skincare ingredient lists)
Domain: GoDaddy

Challenges we ran into

Ingredient Filtering: Ensuring products match the user's needs while keeping recommendations budget friendly required refining our algorithm parameters and determining how to best feed information from the AI to JSON to the algorithm and back to the AI

Balancing Seriousness and Fun: The Looksmaxx was a meme addition, but we wanted to make sure it remained lighthearted and ethical so as not to reinforce a negative self-image

Design: Creating a clean design that matches the track and follows a skincare centric aesthetic required research into deeper topics like how colors and shapes influence mood and perception

AI Integration: Initially, we tried using existing AI models that were pre-trained and tuned for specifically skin-ailment determinations, but realized that the code was no longer functional and relied on deprecated libraries/outdated dependencies. We adapted and turned it into a learning experience, opting to use a more general model from other sources and integrating it into our workflow.

Accomplishments that we're proud of

  1. Fully Automated Routine Generation: Users receive a full custom skincare routine without needing to do manual research

  2. AI-Powered Skin Type and Concern Analysis: We managed to successfully integrate DeepSeek's R1 model into our workflow, tweaking them to fit our niche use case

  3. Budget Friendliness: Unlike other skincare platforms, our focus on affordable and effective products leads to a promising project that has future expansion potential

What we learned

  1. Skincare concerns are highly subjective and individual, so we had to balance AI-driven suggestions with user flexibility

  2. Ingredient-based recommendations proved to be a challenge, but they offer better transparency and user trust than name brand-based recommendations

  3. Determine which tools to use and when - when we couldn't get other existing pre-tuned models for skin-based determinations to work, we adapted and integrated more broad models, such as DeepSeek R1

What's next for Luvera

  1. Personalized User Experience: Build a backend system that allows users to create accounts, log in, and track their skincare routines and analyses over time. Users will also be able to fine-tune recommendations by excluding ingredients they dislike or are allergic to, ensuring a fully customized experience.

  2. More Data Sources: Expand product databases beyond EWG and INCIDecoder to generate even more refined results

  3. Routine Adjustments: Implement feedback-based ML to refine routine recommendations for users over time

  4. Gamification & Community: Introducing features like streaks, achievements, and skincare challenges to keep users engaged.

Built With

Share this project:

Updates