How do you evaluate the readiness of a support ticket classification system for deployment?
Machine Learning / AI
Besides F1 and recall, what metrics were used to evaluate the classifier?
Have you had tabular models in production?
What is a topic in Kafka?
What metrics would you monitor after deploying a model in production: service, data, and model metrics?
Describe your experience with Docker: how to package and deploy an ML service?
There is a 5% lead closing rate, several countries, and agents. How can ML determine which agent to assign a new lead?
What metrics would you use to evaluate a lead assignment system to agents?
Tell us what you know about search, ranking, recommendation, and personalization tasks.
What is a partition in Kafka?
Tell us about yourself and your experience as an ML engineer.
How to normalize the score so that a new agent without history also receives leads and can sell?
How would you explain this solution to stakeholders without statistical training?
Why do you suggest CatBoost for this task?
How would you rate your experience in tabular ML compared to NLP and LLM?
Agent A closed 7 out of 70 leads, and agent B — 2 out of 3. How to compare them fairly?
Which baseline and final model candidate will you choose for this task?
Tell about a case when a model worked well offline but poorly in production: what happened and what did you do?
What is your experience with Microsoft Azure and can you deploy a solution there?
Tell us about yourself in English: briefly describe your most important skills and experience.