Built a simplified pharma research assistant by having three AIs review papers sequentially over two days

Some background first. Over the last two days, I saw a paper AstraZeneca posted on arXiv about using multi-agent systems for drug discovery assistants. I checked it out, and the core idea is evolving from single agents to multiple agents collaborating with divided labor, running on Amazon Bedrock, shared across different departments like clinical, regulatory, and patient support. My biggest takeaway after reading it: if pharma companies are doing this, our lab could actually use a simplified version too.