Lessons learned building AI agents for bioinformatics
Cameron Lloyd · ClawBio
"Every time I open social media, someone's posted about AI agents. I don't know where to start."
Social media → YouTube → documentation → repeat
Always reading. Always about to start.
Stop scrolling and start building.
Scientist by training → bioinformatician & data engineer
Advocate for automation, reproducibility, FAIR data
Saw an advert for the first ClawBio hackathon. Skeptical, naturally.
In an hour, I was able to take an idea to an MVP. Something clicked.
So I submitted a PR. Then another. And I haven't stopped slept since.
fine-mapping cell-detection multiqc-reporter labstep struct-predictor protocols-io
reproducibility module conda-lock SKILL.md template PR audit bot security CI workflows
Most biologists don't know ClawBio exists yet.
In a space full of proprietary black boxes, ClawBio is built in the open — every skill, every decision, reviewable by anyone.
Lowering the barrier so anyone can question an approach, flag a result, or say "this doesn't look right." Domain expertise is a contribution.
Users become contributors. Contributors build better tools. Better tools attract more users. That's how communities grow.
ClawBio is the start of a community.
Others are still trying to do this in isolation.
Together, we can move 10× faster — and in the right direction.
/commons/ for helper functions and test fixtures.Real prompt injection attempt via the ClawBio Telegram bot
Further reading: rewrites.bio — by Seqera
First-time contributor? Check the GitHub issues.
Contributions aren't always code. Your domain expertise matters too.
Unsure where to start? Just ask.