Constructor
Technical interview at Constructor
Constructor hiring-manager technical screen: candidate background, then an NLP modeling case for restoring uppercase and punctuation in speech-to-text output. The discussion covers LLM baselines, sequence labeling, tokenization, BERT-style encoders, metrics, data scale, domain shift and later company context around AI shopping agents and product discovery.
9 questions55 minAudio recording available
Аудио и материалы
Аудио скрининга
0:00 / 55:20
Этап 2 из 3ConstructorML Engineer, AI-агенты для покупок2025-08-28 - 2025-09-25
Собеседования в Constructor: ML Engineer, AI-агенты для покупокТехническое собеседование в Constructor
Выводы и как готовиться
- The case is explicitly about modeling, not backend system design: choose the model family, labels, training data and evaluation.
- The strongest solution frames punctuation and capitalization as constrained sequence labeling rather than free-form LLM rewriting.
- The public audio is trimmed to the interview; unrelated private audio after the call is excluded.