Tell about your main tasks, team, and work structure.
Machine Learning / AI
What is the difference between dropout and batch normalization?
How to handle tool call errors in the agent (retry, fallback)?
What are the advantages and disadvantages of using cross-validation?
What will happen if you do not call model.eval() during validation?
What tool was used for structural document chunking?
What types of convolutions do you know (standard, depthwise, separable, dilated, transposed, grouped)? Ideas, advantages, and disadvantages of each.
What problems can arise when using the Sigmoid activation function together with BatchNorm?
What is OWL-ViT?
What types of models do you know for detection tasks (R-CNN family, YOLO, SSD, DETR)?
When does an inverse matrix exist?
What is the purpose of pruning a decision tree?
MVP task: binary classification with class imbalance, data distributed across 10 machines, with a text feature. What approach would you suggest?
What is a video transformer (ViViT, TimeSformer)?
Why is gradient boosting more commonly used in industrial environments than random forests?
What solutions would you suggest when system load increases?
What is ConvNeXt and why did it catch up with ViT in quality?
How do positional embeddings modify Q, K, V?