Research engineer
The one who makes the experiment hold at scale
A research engineer builds and runs the experiments other people have imagined, at a scale where everything breaks, like the stagehand without whom the staging stays on paper.
- Floor 1The Model
- P4Writes the model
- Established titleIt has been in adverts long enough to mean something.
What they answer for
The deliverableAn experiment that runs, replays, and gives the same number twice. If they leave, research slows by a factor measured in months.
The profile
Five activities, scored 0 to 3- ResearchProducing knowledge that does not yet exist.
- BuildShipping a system that runs, deploys and breaks.
- OperateKeeping it in production: cost, incidents, drift, on-call.
- VerifyMeasuring, testing, attacking. Producing a verdict that holds.
- LeadDeciding, persuading, driving adoption, answering to others.
What the work is
The bottleneck
A research idea fits on one page. Making it run across thousands of machines, without half the attempts failing for infrastructure reasons, takes weeks. That bottleneck created the trade, and it is why labs employ two to four of these for every scientist.
What you have to enjoy
What you have to enjoy is particular, and adverts rarely say it: satisfaction here comes from making other people’s work faster, not from signing an idea. Someone who wants to ask the questions will be unhappy; someone who wants to answer them at scale will do very well.
The trap
The trap is how porous the title is. Depending on the lab, a research engineer proposes their own experiments or only executes other people’s, and the gap is a whole career wide. The interview question fits in one sentence: who decides what gets tested next week?
A week in the role
- Experiment code, throwaway and yet rigorous: it decides what you will believe about the result.
- Scaling: what worked on one machine has to work on a thousand.
- Debugging training runs that diverge, often for a reason unrelated to the idea under test.
- Records, filed so that an experiment can be replayed in six months.
Ways in
- From software engineering, especially with a habit of distributed computing.
- From a doctorate, accepting the move from the open question to execution. That is a choice, not a retreat.
- What the role does not require, whatever people assume: publications. They help; they do not decide.
Reading an advert
2 signs- The advert does not say who chooses the experiments.
- That is the only question separating two very different versions of the same title. Ask it before you sign.
- It asks for a doctorate and first-author publications.
- The advert probably describes a scientist post under this title, or it is copying a requirement nobody thought about.