Writing again

After years of building AI platforms behind NDAs and speaking about the outcomes at conferences, I'm going to start writing here.

I care about the messy middle — the part between "we should use AI for this" and "the trial finished six months early." Most of what I've shipped lives in that middle: knowledge graphs over messy literature, patient digital twins that look nothing like real patients, portfolio intelligence that has to survive a BD team's scrutiny on Monday morning.

You'll see notes on:

• How we actually wire up agentic workflows that clinicians will trust
• Where digital twins help, and the three places they mislead
• What I'm learning building QuriousRI and AlphaMeld
• Talks I'm preparing, papers I'm reading, decisions I'm second-guessing

Short posts. No listicles. Occasional diagrams.

If any of this is useful, get in touch.