Inspiration
Teenagers worldwide face confidence issues due to acne, often stemming from dietary triggers that go unnoticed. We wanted to create a solution to empower individuals to identify and manage these triggers, promoting healthier skin and greater self-confidence.
What it does
ClearBite is a personalized AI-powered assistant that helps users discover the dietary causes of their acne. By asking questions about your diet and providing step-by-step plans, ClearBite identifies potential triggers and suggests healthier alternatives to help you achieve clearer skin.
How we built it
*Frontend: Built using React, HTML, and CSS, we designed a clean and minimalist user interface that is intuitive and user-friendly. *AI Model: We fine-tuned the Mistral AI model (with 7 billion parameters) on a custom dataset curated from multiple research papers. These papers highlighted dietary triggers for acne and their possible replacements. Data cleaning and preprocessing were key to creating a high-quality dataset for fine-tuning. *Backend: Using Flask, we developed an API to connect the Mistral AI model with our frontend. This seamless integration allowed users to interact with the AI for tailored advice.
Challenges we ran into
- Resource Constraints: Training a large model like Mistral on Google Colab presented significant challenges, including limited computational resources and time restrictions.
- Data Preparation: Preprocessing research papers into a structured dataset was a complex and time-consuming process, requiring significant manual effort to ensure accuracy and relevance.
- Integration: Connecting the fine-tuned model to the frontend and ensuring smooth functionality was a challenging but rewarding process.
Accomplishments that we're proud of
- Successfully fine-tuned a 7-billion-parameter model on a custom dataset within the constraints of Google Colab.
- Creating a fully functional application, complete with frontend, backend, and an integrated AI model, in just 24 hours.
- Building an intuitive interface that enhances user experience while tackling a sensitive topic like acne with care and precision.
What we learned
- The importance of creating high-quality, structured datasets for fine-tuning large language models.
- Practical experience in fine-tuning models and testing different architectures for optimal performance.
- Insights into frontend-backend integration and how to manage resource limitations effectively.
- How to collaborate effectively under tight deadlines to deliver a polished product.
What's next for ClearBite
- Image Analysis: Integrating AI-powered image detection to analyze the type and severity of acne, enabling more tailored recommendations.
- Expanded Dataset: Adding more diverse dietary datasets to improve model accuracy and inclusivity.
Log in or sign up for Devpost to join the conversation.