AI/ML Interview Hub

What is the bias-variance tradeoff?

Explain precision, recall, and F1-score. When would you prioritize recall over precision?

What is regularization and why is it used?

How does gradient boosting differ from random forests?

Implement k-fold cross-validation from scratch without sklearn.

Implement logistic regression with gradient descent from scratch.

Design a recommendation system for a 10M user e-commerce platform.

Explain vanishing and exploding gradients. How do you fix them?

What is batch normalization and why does it help training?

Explain the difference between CNN, RNN, and Transformer architectures.

Implement a simple neural network in PyTorch for binary classification.

What is TF-IDF and when would you use it over word embeddings?

Explain the attention mechanism in Transformers.

Fine-tune a BERT model for text classification using HuggingFace.

What is RLHF and how is it used to align LLMs?

What is prompt injection and how do you mitigate it?

Build a streaming chat completion with Groq and handle tool calling.

What are the main components of a RAG pipeline and what can go wrong at each stage?

What is the difference between naive RAG and advanced RAG?

Build a basic RAG pipeline using FAISS and sentence-transformers.

What is model drift and how do you detect and handle it?

Design a CI/CD pipeline for ML models.

Design a real-time fraud detection system for 100K transactions/second.

Design a system to serve 1B+ embedding lookups per day for a semantic search engine.