First round: phone conversation with one of the principals--an enjoyable conversation. I was upfront about my inexperience with drug applications; the the interviewer said that was not a problem and they were more interested in MLE expertise.
Second round: an open-ended 'take-home'-style ML code challenge controlled via a private Github repository. It was suggested to take 3-4 hours but timing was flexible (up to 7 days after I activated the Github invitation). I received guidance to focus more on ML Engineering skills than on research, and to focus on engineering practices instead of state-of-the-art results.
The coding challenge was straightforward and interesting, but certainly not feasible for someone without domain knowledge to complete in 3-4 hours.
I received a rejection notice about a week later, with detailed feedback (which in and of itself is a nice positive), but it was clear from the feedback that the criteria and areas of emphasis I was given were very different from the actual evaluation criteria (in particular domain knowledge).
Because of the mismatch between stated and actual evaluation criteria, and the scope of work required by the problem, I would consider this a poor interview experience, which is a shame because I am excited about the company mission and vision. I'd encourage Postera to fix their assessment process. There are some very good Medium and Linkedin articles on the pitfalls of take-homes and best practices for doing them right. For example, If they must use an at-home challenge, evaluation criteria should be clear (or at least consistent), and time limits should be standardized, enforced, and no more than 2 hours.
