Roadzen interview questions

1. Why use rag 2. How sentiment anaylysis model work 3. gradient boosting model, XGBoost model works 4. Random forest work, how data split 5. How gradient boosting work all. Concept 6. Self attention, cross attention, multi head attention, masked multi head attention 7. Temperature what to do in mathematical term in llm 8. What will be the input size in transformer in before softmax step 9. Difference between transformer and LSTM 10. How to remove sparsity, how to handle 11. XGBoost major parameters 12. DBSCAN AND KMEANS SENERIO QUESTION, IF DATASET HAVE CLUSTER and in dataset HAVE TWO OUTLIER, IF DBSCAN APPLY THEN WHAT TO DO DBSCAN AND K MEANS on outlier. 13. WHAT IS DBSCAN, KMEANS AND H CLUATRING I DEPTH 14. How to decide which vector db used from faiss, chromadb etc. 15. Docker file, docker compose file 16. Fast api, can handle concrently menas multiple request at a time 17. Why positional encoding required, while already index posion have 18. Encoding method word2vec working Procedure 19. Types of chucking