About · Justin H. Johnson
Two decades in biotech. I went from the lab to running large data and AI teams. But that's not who I am. I'm a builder. Always have been.
VP, Head of Data and Applied AI, R&D at Tempussince 2026-08-03
Receipts: I'm Justin Johnson. I Build Things. · Twenty-five years and half a mile
The web in the mid-nineties. Genomics in the early two-thousands. Clinical sequencing around 2010. Agentic AI from about 2023. Each time I arrived while the field was still arguing about what it was.
The address changed. And that's about it.
VP, Head of Data and Applied AI, R&D at Tempus, since 2026-08-03.
Drag the route or use the arrow keys. Dates from my post “Twenty-five years and half a mile”, 2026-08-11.
The morning before my first day at Tempus I went for a run through Chicago.
A couple of miles in, I recognised the route. Not the way you recognise a place in a photograph. The way your legs recognise a hill before you do. I had run this before.
I ended up standing on a sidewalk at seven in the morning doing arithmetic I had not planned on. It came out at twenty-five years and half a mile. That is the distance from the web applications company where I interned in college to the building I walked into a few hours later as a VP.
I got the goosebumps. Then I finished the run.
Receipts: Twenty-five years and half a mile · The film: Half a Mile (poster, coming)
Thirteen years at AstraZeneca. I joined to work on genomics infrastructure and left running the AI Centre of Excellence for R&D.
The work was FAIR data foundations, a prioritisation layer, agentic systems that did real research tasks, and a governance pathway that took model provisioning from weeks to hours. That last one mattered most and got the least attention, which is normal. I ended up repeating a line often enough that people started saying it back to me.
Safe is the fast way.
None of those four platforms was assigned to me.
So I build at night. Weekends. Early mornings before anyone's awake. On planes. On trains. In hotel rooms.
Not because I have to. Because I can't not.
One I'm proud of: I built ARIA, a research agent that runs on my own hardware, generating and testing hypotheses around the clock. It heals its own crashes. It is patent pending, and the paper is out (DOI 10.5281/zenodo.21383332).
I fired up Claude Code and said, let's start thinking about some experiments we could set up. And it said, oh, by the way, did you know that ARIA's already done this?
I called myself the second mover in my own project.
Write the first version. Prove it works. Hand it over. This time the hand-off was to a machine, and it is on the record.
Week one. 278 commits, 96 of them mine. The first version is always mine. That is not pride, it is how I find out whether the thing can exist.
Week three, 146 of 1,707. The last week I did real work inside it. Proved it ran.
Week four: four commits of mine against 1,894. The hand is open.
Three weeks, 5,196 commits, none of them mine. The baton is hers. The run kept going while the building was empty.
Fifty days. 10,557 commits on this one instance; 321 crashes, 314 healed without a person. I built it, proved it, and got out of its way.
Data: weekly commits from the ARIA paper, Figure 7 (the aria-se run, 2026-02-06 to 2026-03-28); recovery events from its evidence files. The four platforms at AstraZeneca followed the same method; their numbers are not public and are not shown. Read the paper
Receipts: Navari · AIthena · Agentis · (AI)Xplore · ARIA paper, DOI 10.5281/zenodo.21383332 · The Word I Didn't Write
Genomics was going to give us personalised medicine within a decade of the genome being published. It was oversold on timing and roughly right on direction. I have spent twenty-five years running into the absence of the data it needed.
At TIGR the software to assemble and annotate a genome at that scale did not exist as a thing you could buy, so you wrote it. I helped computationally assemble the first human genome, the first large-scale metagenome, and the first synthetic genome ever made.
At EdgeBio I built the sequencing and bioinformatic interpretation platform for a CLIA lab. There was no reference architecture, so we built one, and most of it was wrong in ways we only found by running it against real samples.
From genomics infrastructure to the AI Centre of Excellence for R&D. What was young was the fit between AI and a hundred-and-fifty-year-old company.
Molecular, imaging and clinical outcomes on the same patients, at a size a model can learn something from. I would like to find out what is on the other side of it.
On the side · SocialEyes
A team in Nepal building AI that screens for disease from a single eye scan, in clinics with no specialist down the hall.
We pointed ARIA at real work for SocialEyes. It helped their small team move at a pace their headcount shouldn't have allowed. Something I built for myself ended up helping people I'd never otherwise have reached. Then we won an NVIDIA Inception grant to carry the work forward.
The interesting thing about the eye is it's the one place that you can see the blood vessels and the nerves without cutting people open.
Taking frontier AI like this and pointing it at people that a lab can't touch. The people will never hear her name, and that's fine.
Receipts: Twenty-five years and half a mile · I'm Justin Johnson. I Build Things. · Episode 100, from 5:05
Domino Data Science Leaders · Episode 100 · 2026-08-24 · 30:04
Did you know that ARIA's already done this?
I called myself the second mover in my own project.
The one place that you can see the blood vessels and the nerves without cutting people open.
I came to coding sideways through biology. Science was kind of that passion. Code was kind of how I got there.
You can't read your way across this. The crossing is personal.
My own arc has gone from really heavy steering of these things to just trying to stay out of their way, so I'm not slowing them down.
Get uncomfortable, build, get out of the way of AI.
Quotes from the episode's auto-generated captions, lightly cleaned of filler. Each stamp starts the episode at that second.
Receipts: YouTube: Becoming the builder-conductor · Apple Podcasts
The book · paperback 2026-07-01
The builder and the leader stopped being two different people.
The skills you build collaborating with AI at night, intent, orchestration, judgment, knowing when to step in and when to stay out, are the same skills that make you better at leading people during the day. So I wrote it down.
Buy on Amazonbuilder-leader.com
You can't read your way across this. You can't buy your way across it. The crossing is personal, because somebody's got to do it, and then somebody's got to teach somebody else how to do it.
Posts: post index, as of 2026-10-08 (2026 Q4 is eight days).
Receipts: builder-leader.com · Amazon · Run Data Run · AIXplore
People keep telling me this is a full-circle story. I do not think it is. A circle means you came back to where you started, and I never left.
I have had one job for twenty-five years, which is to turn up somewhere the rules have not been written yet and start writing them badly until they get better.
Get uncomfortable, build, get out of the way of AI, and you will end up in places you may not have expected.
Receipts: LinkedIn · GitHub BioInfo