Domande di colloquio per Algorithm Engineer, Large Language Model presso Shopee

Technical Assessment (20 minutes) LeetCode Question: Longest Palindromic Substring Coding live where the interviewer can see your code on a shared editor. Discussion and Questions (40 minutes) Self-Introduction Project Discussion: Explain a project related to LLM LLM Creation: How is a Large Language Model created? Model Architectures: Explain the different models and their architectures (e.g., GPT, LLAMA, Falcon, BLOOM) Positional Embeddings: What is the purpose of positional embeddings? Normalization Techniques: Difference between BatchNorm and LayerNorm Retrieval-Augmented Generation (RAG): - What is RAG? Explain its purpose. - Even if you use RAG, there are hallucinations occurring. Why is this so and what can you do to mitigate? - What other current academic advancements in RAG? Explain about some frameworks that are trending. Scenario Question: If we have large-scale data (billions of records) from the web, how can I efficiently select math-related data? Discuss using distributed computing frameworks if possible.