Getting hired
What we read for, and how to prepare.
Every stage of the process looks for something specific. These are the three we read most closely, followed by the questions applicants ask us most.
Writing good code
On a platform where a bug can reach a patient record, good code has a specific shape.
It is schema-first, with Zod. It returns errors as values and does not throw them. Its ports and adapters stay true to what they claim, so a tool can be tested on its own. And it is legible to the next person, because we read far more code than we write, and the best change is often a removal.
Some of what we read in an application is not the answer but the code: whether it is correct where it matters, clear to whoever comes next, and open about what it depends on, and whether its author reduces before adding. This runs through every stage. There is no separate test for it.
How to prepare
Show us code you are willing to put your name on, on GitHub where we can read it. A real thing you built tells us more than an exercise written for an interview.
Working inside enforced systems
Every Pensieve tool reaches the data through the core systems, and through nothing else.
An automated check across the whole graph fails the build when a tool tries to go around a core system. We talk with you about working inside constraints like that: whether a boundary reads to you as the scaffolding that keeps the whole thing correct, or as friction to route around.
The question underneath is whether you build well inside a system you did not design, respect a boundary that exists for a reason, and choose the difficult correct change over the easy quick one. It comes up in conversation, and nothing about it is timed.
How to prepare
Think of a time you worked inside a real constraint, such as a platform, a protocol or a safety requirement, and made something good because of it. Then find a small bug in an open-source project and fix it, and notice how much of the time went into finding the right place to change.
Open-ended questions
Part of every conversation has no right answer.
We may ask about your views on the world, the social structures you believe in or oppose and why, art, or medicine and the practice of it: anything outside the radar of credentials. It is how we learn whether we think well together when a question is genuinely open.
We are listening for how you reason without a rubric, whether you can hold a real position and change it when a better one arrives, and the shape of your judgement when nothing is certain.
How to prepare
There is nothing to rehearse. Come as you actually think: a rehearsed answer is worth nothing here, and a real one is worth a great deal.