Machine Learning / AI

What is the difference between U-Net and FPN approaches in segmentation?

Do you specialize in computer vision or are you willing to work in other domains, such as NLP?

What methods exist for reducing the size of a model while maintaining quality?

What are the methods of text processing in NLP from simple to modern neural networks?

Middle

СБЕРМагнит

What is RNN-T (transducer) and its advantage for streaming?

What is text normalization (NFC, NFKC) and why should it be done before tokenization?

What are the pitfalls when building a knowledge graph from unstructured text?

In which cases is it worth applying a two-step detector in production?

What is a receptive field?

What will happen if you do not call model.eval() during validation?

What is the CLIP architecture?

What is FastAPI?

What is the difference between GET and POST requests in HTTP?

What helps U-Net to preserve information at low-level features?

What is gradient accumulation and why is it needed on limited memory?

What is the problem with fine-tuning CLIP?

What is model-based RL and what are its advantages over model-free?

Middle — Senior

МагнитГазпромбанк

What is replication in ClickHouse?

What is the architecture of ViT (Vision Transformer)?

What is supervised fine-tuning (SFT)? What data is needed?

Middle — Senior

ТензорВижнТехМагнит