So I need help training a LLM to help students prepare for their emergency medicine board examinations. My idea is this, we will use Stanford's openly available BioMedLM from Huggingface as the foundation model. We shall add emergency medicine-specific knowledge as embeddings, I have pdf's of the knowledge base. We shall also scrap a few websites that have FOAMed (Free Open Access Medical education ) and add these. Next, we'll create a dataset for fine-tuning the model. This dataset will include multiple-choice questions from written board exams and oral questions from oral board exams. If the model struggles with handling both types of questions, we can even build separate models for each. Fine-tuned model to be hosted and fine-tuned on Google Cloud using their free credits for start-ups. As a domain expert, I'll handle the subsequent prompt engineering. This will be made freely available to students in my class. Because of the free resources utilized under the Creative Commons license, it will not be a commercial project so I want to keep costs low. That's the reason for trying to get this done by freelancers rather than software companies.
This job is already closed and no longer accepting applicants, sorry.