Technical interview at Tochka

Выводы и как готовиться

• The interviewer probes whether production ML answers include decision costs, labels, threshold ownership and post-launch feedback, not only model choice.
• The theory block is broad but coherent: metrics, linear-model interpretability, boosting internals, Transformers, gradient stability and optimizers.
• The coding task is intentionally simple, but it catches edge cases around zero, negative numbers and zero-padding inner thousands groups.
• The public audio is trimmed to the interview portion; a short unrelated tail after the call is excluded.