Online Test consist of 13 questions out of which 11 were MCQ and 2 were ML programming. Wine quality prediction with more than 70% accuracy was the criteria in one program and in another room occupany only using knn was the criteria. The platform DoSelect is the worst, you are expected to complete 2 programs with good accuracy and mcq in just 90 minutes. Worst experience with DoSelect and Fractal question patterns

The interview was primarily focused on Generative AI concepts and their practical applications. Questions covered Large Language Models (LLMs), prompt engineering, Retrieval-Augmented Generation (RAG), vector databases, embeddings, fine-tuning techniques, and AI agent frameworks. The interviewer also asked scenario-based questions on designing GenAI solutions, selecting appropriate architectures for different use cases, handling hallucinations, evaluating model performance, and addressing scalability, security, and deployment challenges. Additionally, there were discussions around real-world implementation of GenAI applications, model optimization, and best practices for building production-ready AI systems.