SaRAH
Introducing
is revolutionary AI skin routine creator that is designed to provide personalized and effective routines in a matter of seconds. SARAH (Simplified Assessment and Recommendation for Achieving Health) has been trained using thousands of clinically validated routines to ensure that it can recommend the best routine for your unique skin with precision and confidence.
Art by Laura de Vazeilles
The model analyzes data from a form filled out by the user (e.g. skin type, skin conditions, body area) and provides personalized skin routines to improve overall skin health and well-being. It identifies patterns and correlations within the data to make more accurate assessments and recommendations. SaRAH is a generative pre-trained model that does not learn from use and does not save personal data.
Motivation
Neglecting your skin can have serious consequences for your health. Skin is the body's largest organ and acts as a barrier against harmful external elements such as bacteria and viruses. When skin is not cared for properly, it can become more susceptible to infections, diseases, and even skin cancer.
Factors such as geographic location, socioeconomic status, and cultural differences can limit people's access to quality skincare products and services. In addition, misinformation and lack of education about skincare can lead to ineffective or harmful practices. This is why we want to expand SaRAH's capabilities in order to help people worldwide to take care of their unique skins.
SaRAH Limitations
Lack of complete understanding: While SaRAH can analyze data and provide personalized recommendations, it may not have a complete understanding of all factors that could be impacting an individual's skin health. This means that some recommendations may not be as effective as others.​
Limited by training data: The effectiveness of SaRAH depends on the quality and diversity of the training data it has received. It may not be able to provide accurate recommendations for a diverse range of skin types and conditions.
Inability to account for unique circumstances: There may be unique circumstances that an AI model cannot account for, such as underlying health conditions, environmental factors, or medication use, which could impact the effectiveness of recommended skincare routines.​
Dependence on user input: The accuracy of the AI model's recommendations relies heavily on the accuracy and completeness of the data provided by the user. If the user does not provide complete or accurate information, the recommendations may not be effective.​
Behind the UI
SaRAH is a fine-tuned version of the GPT-3 model that was built using the TensorFlow open-source software library. TensorFlow is a framework for developing AI models that allows to build and train models efficiently. A TF model is created by defining the network architecture using a high-level programming language, in this case Python. The model is then trained on a dataset using various optimization algorithms to adjust the weights of the network, minimizing the error between the predicted and actual outputs.
GPT-3 is a state-of-the-art language model that has been pre-trained on a massive amount of text data using unsupervised learning techniques. However, the pre-trained model may not perform optimally for specific tasks or applications, as it may not have been trained on the specific domain or with specific data. This is where fine-tuning comes into play.
Fine-tuning is a technique where the pre-trained GPT-3 model is further trained on a smaller dataset that is specific to a particular task or domain. The fine-tuning process involves feeding the model with the specific data and fine-tuning the model's parameters to optimize its performance on the given task.
Once the fine-tuning is complete, the resulting model can be used for various news applications. The fine-tuned model can be deployed in production environments and used to generate outputs or make predictions.
CHECK OUT OPEN AI's GPT
Hosted and built with Streamlit
Streamlit is an open-source Python library that allows developers to easily create web applications for data science and machine learning projects. It enables developers to create interactive, real-time web applications without getting involved in difficult web development or JavaScript. ​
