Probably can't actually make money with this, but:

Train a simple random forest or even use document embeddings plus transformer on whatever structured clinical trial data points to predict drug success and then adjust rnpv or predict market size or sales

When a drug gets approved by the FDA or when a new drug is filed, scrape this data and identify the closest drug in similarity using LLM embeddings plus cosine similarity and invest in the closest drug

trade based on announcement sentiment using fine tuned bert classifier… is within minutes good enough?

> quants are already doing this and playing w/ volume etc.

Predict clinical trial success:

• - https://pubmed.ncbi.nlm.nih.gov/37483175/

• - https://www.sciencedirect.com/science/article/pii/S2666389922000186?pes=vor

• - https://arxiv.org/html/2402.06512v3

• 
  

Predict clinical trial failures/termination:

• - https://www.nature.com/articles/s41598-021-82840-x

• - https://github.com/futianfan/clinical-trial-outcome-prediction/tree/main

Predict whether or not clinical trial will recruit successfully:

• - https://arxiv.org/html/2407.13115v1