Technical: math, CS, ML theory, production and superapp ML System Design

Broad oral round with probability, statistics, Python/CS basics, metrics, overfitting, trees, drift, Transformers, A/B tests, production ML and superapp recommendation system design.

13 questions1 case1 hr 46 minAudio recording available

Question languageРусский English Оригинал

Пройти собеседование Весь ход интервью

Аудио и материалы

Аудио собеседования

0:00 / 1:46:26

Этап 2 из 4inDriveSenior Data Scientist, Geo Search2025-09-19 - 2025-10-22

Собеседования в inDrive: поиск по геоданным

Техническое собеседование в inDrive

Этап 1 Весь процесс Этап 3

Выводы и как готовиться

The technical stage was breadth-heavy rather than coding-heavy: many short oral probes plus one ML System Design case.
The interviewers cared about practical ML judgment: imbalanced metrics, overfitting, drift, A/B test interference and production latency tradeoffs.
The superapp case exposed a key training point: move from vague embedding intuition to a scored constrained ranking objective.

Technical: math, CS, ML theory, production and superapp ML System Design

Аудио и материалы

Аудио собеседования

Intro and candidate project overview

Conditional probability: cats heavier than 5 kg

Law of large numbers, CLT, and the sample-mean distribution

Operation complexity for linked lists, Python list, dict, and strings

Precision, recall, and ROC-AUC with strong class imbalance

Overfitting, regularization, and hyperparameter selection

Bagging, boosting, and tree regularization

Data drift and production model monitoring

Attention, positional embeddings, BPE, and LoRA

A/B testing with marketplace network effects

Production ML serving and latency optimization

Python multithreading vs multiprocessing

Recommendations between verticals in a superapp

Вопросы кандидата про R&D growth team, Solution Delivery и следующие этапы

Выводы и как готовиться