PDT Partners Machine Learning Engineer Interview Experience — Project Deep Dive, Then a Find-Divisors Follow-up

PDT Partners·Machine Learning Engineer·Aug 2026
Technical ScreenWithdrewmedium

I interviewed with Citadel Securities (see my previous post) and PDT back to back. I originally wanted to post both back to back too, but something came up and I got delayed.

The PDT team is different from the Citadel one. The link they sent was for a research engineer role, but both the headhunter and HR said it would be an LLM team, and it was hybrid 3 days a week, so I was really excited. But because it meant moving to New York, and I was busy with other things at the time, I only did the three technical rounds and did not continue to the next stage. HR said they would not hold the result for me.

Overall the PDT Partners experience was very good, but I get the feeling that, at least for this role, the interview process had not been fixed yet at the time (June). It was a bit of a mixed bag.

Today I will just talk about the first round. Before the interview I was told it would be coding, but the interviewer led me through a 15-minute project deep dive first. I checked in the middle whether I was supposed to go into details, and got a yes.

For the deep dive he picked the first industry project on my resume, which was honestly pretty old. It was a small LSTM model but with large-scale inference. He asked a lot of infra details, including what compute it was deployed on, the latency requirements, and what it was used for. On the model side he focused on how to reduce latency, and I said quantization. He asked why we didn't use a transformer, which was already mainstream back then, and I answered memory constraints, plus that in testing it showed no advantage at typical sequence lengths. He asked whether I would make the same choice if it were this year. I did not answer that well; I said I would start from experiments, and he did not follow up.

The coding looked very simple at first: find divisors. I finished it quickly, in O(sqrt N) time. Then he asked whether it could be done in better than sqrt N. He gave a lot of hints, which I honestly could not follow or remember, and I still did not get it. You can find some discussions about it on Reddit if you want a reference.

In the end I passed (the results of this round came together with the other two technical rounds). Maybe this last bit of extreme pushing was just the interviewer's personal hobby.

Published

Curated and edited by PracHub

Practice the questions from this interview

Discussion

Sign in to join the discussion. The author is notified of every comment.

Loading comments…

Interview at a glance

Company
PDT Partners
Role
Machine Learning Engineer
Rounds
Technical Screen
Outcome
Withdrew
Difficulty
medium
Interview date
Aug 2026
Questions from this interview
2 questions

Real PDT Partners interview experiences

First-hand reports from PDT Partners candidates — the rounds, the questions they were asked, and how it went.

All 6 PDT Partners interview experiences