Entendre Éloquente
Inspiration
The skincare world is full of advice, buzzwords, and endless product options—yet many people feel lost when building routines that actually work for their unique skin types and concerns. We wanted to build something that cuts through the clutter, giving users smart, tailored skincare help—just like talking to a trusted expert who actually understands your skin. This project was inspired by a mix of personal struggles with skincare, the growing demand for personalized wellness, and the exciting capabilities of AI to deliver expert-like assistance.
What it does
Entendre Éloquente is an AI-powered skincare assistant that:
- Builds customized morning, evening, and night skincare routines based on user input.
- Recommends products based on skin concerns (e.g., acne, dryness, hyperpigmentation).
- Responds to general skincare queries, such as ingredient clarifications, routine order, or safe combinations.
Users can interact naturally, and the AI adapts recommendations based on context, preferences, and evolving skincare needs.
How we built it
- Natural Language Understanding: We fine-tuned an LLM to understand skincare terminology and context using domain-specific data and expert articles.
- Routine Engine: We created a rule-based framework combined with intent classification to generate routines in a structured way (AM/PM/Night).
- Product Recommender: A hybrid model was used—keyword mapping for active ingredients + similarity scoring using embeddings trained on product descriptions.
- Frontend: A simple UI was designed using Streamlit to make it accessible for non-technical users.
Challenges we ran into
- Data reliability: Skincare data online is fragmented and often conflicting. We had to carefully curate sources to avoid spreading misinformation.
- Personalization: Skin types and sensitivities vary widely. We worked to balance general advice with personalization logic.
- Routine logic: Creating routines that respect ingredient conflicts (e.g., retinol + AHAs) and layering order was trickier than expected.
- Tone and trust: Ensuring the AI’s language felt supportive, safe, and not overly clinical was an iterative process.
Accomplishments that we're proud of
- Built a system that gives context-aware skincare routines based on natural conversation.
- Developed a functional prototype that could recommend real products, not just generic ingredients.
- Embedded a skincare logic engine that respects dermatological principles like ingredient synergy and order of application.
- Maintained user-friendly, informative responses without sounding like a textbook.
What we learned
- AI agents can be powerful in niche, high-context domains when paired with domain rules and quality data.
- User trust is crucial in wellness applications—transparency and gentle language matter.
- The importance of explainability—users don’t just want answers; they want to understand why something is recommended.
- Building conversational UX takes more effort than raw functionality—it’s about guiding and educating, not just replying.
What's next for Entendre Éloquente
- Feedback learning loop: Incorporate user feedback to refine recommendations over time.
- Skin image analysis: Integrate vision models to detect skin issues visually and enhance personalization (collaborative).
- Product filters: Add constraints like budget, cruelty-free, or fragrance-free to match user values.
- Mobile app: Build a mobile-first version with reminders, tracking, and routine journaling.
- Dermatologist collaboration: Partner with professionals to verify logic and add credibility.
Entendre Éloquente is more than a skincare bot—it's a step toward empowering users with AI-driven, personalized wellness guidance.
Log in or sign up for Devpost to join the conversation.