Machine Learning / AI

How do mean/max pooling differ from learnable pooling?

How to choose the size of embedding vectors?

What is the RULER benchmark for long-context?

What are shadow deployment and canary deployment?

Middle — Senior

2gisНейроЛаб

What are the limitations of Whisper and how to bypass them (long-form, hallucinations)?

Middle — Senior

WildberriesНейроЛаб

What is NSP (next sentence prediction) and why did RoBERTa abandon it?

Middle — Senior

WildberriesИннотехНейроЛаб

How does FlashAttention v2 differ from v1?

What is LoRA and why does it work (low-rank hypothesis)?

What is meta-prompting and what are its use cases?

What is cross-encoder QA and what is the input format?

Why is bias=False used in Linear layers in modern LLMs?

Спроектируйте систему голосового ассистента с low-latency (<500мс) end-to-end.

Senior

X5 techНейроЛаб

What compression formats (binary, scalar quant) are available in VDB?

What is beam search and why is it rarely used for open-ended generation?

What are the problems of BERT in text generation tasks?

What is quantization and what schemes are there (PTQ, QAT, AWQ, GPTQ, GGUF)?

What is the role of payload filters in a vector database?

What is DocVQA and what models are popular there?

What is a filter bubble and how is it detected?

What is the idea of the Byte Pair Encoding (BPE) algorithm? Describe the training steps.

Junior — Middle

VK CloudНейроЛаб