FitFiction

From Insight to Action: Navigating Your Health Odyssey with Precision and Play Get personalised health insights, interactive advice, and data-driven paths to a happier, healthier you

Comment

FitFiction

Inspiration

The inception of FitFiction was sparked by the observation that while there's a wealth of health-related data available to individuals, deciphering this data into actionable, personalized insights remains a challenge for many. We envisioned a platform that not only simplifies this translation but does so in an engaging, user-friendly manner. Our goal was to blend the precision of health metrics with the allure of interactive technology to motivate users toward healthier lifestyles.

What it does

FitFiction is a comprehensive digital health platform that offers stress level predictions, personalized health recommendations, and an interactive chat interface powered by our LLaMA model trained on health data. Users input health-related parameters, and FitFiction provides insights and actionable advice. Additionally, our platform features data visualization tools, allowing users to see how their health stats compare with others, fostering a deeper understanding of their wellness journey.

How we built it

FitFiction combines popular libraries such as React, Flask, and Python to provide a seamless user experience, leveraging data science and machine learning for its core functionalities. Our platform uses predictive model based on ridge regression, designed to forecast stress levels from user inputs with high precision. At the heart of our platform is the health recommendation engine, enriched by the integration of the LLaMA 2 model. This cutting-edge AI, with its 13 billion parameters, harnesses the power of PyTorch for deep learning, enabling us to push the boundaries of personalized health advice.

Our LLaMA 2 model, specifically trained on a meticulously curated corpus of thirty health documents, was operationalized using LangChain. This innovative framework significantly simplifies the integration of large language models into applications, allowing us to offer an interactive chat interface. Through this interface, users receive tailored health recommendations and engage in meaningful conversations about their wellness journey. FitFiction not only showcases the potential of combining Generative AI and health but also reflects our dedication to offering actionable, personalized health insights.

Challenges we ran into

One of our primary hurdles was operating within the bounds of limited computational resources to train the LLaMA model effectively. Optimizing the model to ensure both efficiency and accuracy, given the resource constraints, required innovative solutions and optimizations. Additionally, sourcing a high-quality, relevant corpus to train our LLaMA model on presented its own set of challenges. Identifying, curating, and preprocessing thirty health documents to form a coherent, informative base for training took considerable effort, underscoring the importance of quality data in machine learning. These challenges not only tested our technical prowess but also our resourcefulness in overcoming constraints.

Accomplishments that we're proud of

Successfully creating a user-friendly platform that demystifies health data into actionable insights stands as our proudest achievement. Furthermore, our ability to integrate advanced machine learning models into a responsive web application marks a significant technical accomplishment.

What we learned

Throughout the development of FitFiction, we deepened our understanding of machine learning, generative AI, and web development. We also gained insights into the importance of user experience design in health technology. Bridging the gap between complex data analysis and user-friendly interfaces was a valuable lesson in the importance of interdisciplinary collaboration. We also learned about different approaches that we can leverage to overcome computational bottleneck when training large language models like LLaMA 2.

What's next for FitFiction

Looking ahead, we aim to expand FitFiction's capabilities by introducing simulation features that allow users to visualize potential health outcomes based on hypothetical lifestyle changes. Enhancing the LLaMA model for broader health topics and incorporating real-time data from wearable devices are also on our roadmap. Our vision is to make FitFiction a holistic wellness companion, guiding users not just toward understanding their current health status but toward envisioning and achieving their optimal well-being.

Built With

Share this project:

Updates