Gazal.Ai, Pioneering Open-Access in Healthcare Knowledge with AI

In a groundbreaking move aimed at democratizing healthcare knowledge, Tachyhealth, has launched Gazal.Ai, an open-access large language model (LLM) focused on healthcare and medical domains. This state-of-the-art AI solution is developed to enhance the accessibility of health information, facilitating better care and fostering innovation in healthcare. Gazal.Ai was developed by leveraging the robust framework of Mixtral, a Mixture of Experts (MoE) Model, encompassing an impressive 8x7 billion parameters, making it one of the most advanced (yet less complex) models in the field of healthcare AI. This model is tailored to understand and process complex healthcare language, ensuring that it can serve a wide array of healthcare professionals and researchers.

Objectives and Impact:

The primary goal of Gazal.Ai is to expand access to healthcare knowledge, breaking down the barriers that often hinder the flow of information in the healthcare sector by providing an open-access platform.

Technology and Features:

Gazal.Ai’s architecture is built on the latest advancements in AI and machine learning. The model's 8x7 billion parameters allow it to process and generate information with an unprecedented level of accuracy and detail.

Key features include:

• Complex Query Handling: Gazal.Ai can understand and respond to intricate medical queries, making it an invaluable tool for diagnosing complex cases.
• Multilingual Capabilities: To ensure distribution of accessibility, Gazal.Ai supports multiple languages, allowing it to serve a global audience with different languages including Arabic with diffrent dialects.
• Continuous Learning: Gazal.Ai is designed mainly to learn continuously from new data, ensuring it remains up-to-date with the latest medical research and clinical best practices.

Experimental and technical details:

The experiments have been performed on Vertex AI with A100 GPUs. For simplicity, the finetuning was adjusted to be on 5 epochs which took in total 215 hours. Gazal.Ai characteristics are given in Table 1.

Results and comparisons:

To validate the results, we used the lm-evaluation-harness metrics to validated the performance of Gazal.Ai on some popular medical multichoice and question\ answering datasets. There is no overlap between training and testing Gazal.Ai performance. A comprehensive comparison between Gazal.Ai and other medical pretrained models including the most recently published Med42 beside Clinical Camel and GPT-3.5 is given in Table 2.

As seen in Table 1, Gazal.Ai outperformed other clinical pretrained model in 3 of 7 datasets, given the fact that we have no further information about the experimental details such as number of epochs which was set to 5 in our trainings. We are working on a new version that is currently under training. We expect this version to surpass the benchmark in several parameters.

Access to our model:

To use Gazal.Ai, it’s available on Tachyhealth Huggingface page at: https://huggingface.co/TachyHealth/Mistral-8x7b-Medical-Finetune_QA_MCQ_Gazal_Ai .

Conclusion: Gazal.Ai represents a significant step forward in the integration of AI into healthcare. By making advanced medical knowledge widely accessible, Gazal.Ai not only supports healthcare professionals but also empowers patients, ultimately contributing to improved healthcare outcomes globally. With its sophisticated technology and noble mission, Gazal.Ai is poised to transform the landscape of medical information and education. Achieving NLP tasks by leveraging LLMs in the medical field, especially for QA and multichoice use cases, presents significant challenges. However, that opens new eras in applying LLMs in more use cases to cover more challenging areas in the healthcare domain.

About TachyHealth:

TachyHealth is pioneering the next generation of healthcare solutions by leveraging artificial intelligence to drive value-based healthcare. Our technology suite includes advanced tools in Artificial Intelligence, Machine Learning, Neural Networks, and Deep Learning, aimed at enhancing the efficiency of healthcare players at an enterprise scale.

Legal Disclaimer:

The information provided by TachyHealth, Inc. on all platforms is for general informational purposes only. All information is provided in good faith; however, we make no representation or warranty of any kind, express or implied, regarding the accuracy, reliability, or completeness of any information. Privacy Policy: At TachyHealth, your privacy is a top priority. We adhere strictly to legal standards for data protection. For detailed information, please visit our privacy policy on our website.

Contact Us:

For more information about our services or to get in touch with our team, please visit www.tachyhealth.com or contact us at [email protected]