Explain projects, why RAG was used in your project, how was agentic architecture was setup, why and how vector db are indexed, how will you find most recurring topic of customer grievance from set of tech support scripts, what is temperature in LLM and what does it exactly do to modify output personality of a LLM. Programming: find second highest element in a list
Leetcode 1854 . Only had time for this one.
1. First round: Basic questions on Projects Tech stack I used Technical questions on LLM, NLP Differences between Transformers, LSTM and RNN 2. Second Round: Have you heard of the Vanishing gradient issue? What is called a Dropout layer? How would you evaluate the transfer model? Can you name some parmenter? What technique would you use? What is called " perplexity" and "bleu score" in LLM? Do you have experience working with LSTM and CNN or RNN? Have you worked with GAN? What is called LLM Beam search? check other searches What is called context based recommender? What is called self attention mechanism? Ans LSTM What is the position encoding numbering in context on LLM? (Transfer modeling) What is called overfitting and underfitting? (Who would you overcome the scenario) What is called supervised and unSupervised learning? (Example) Have you worked with reinforcement learning? Explain Transformers Scenario based Design AI based system using Python to address the following: 1. Code Styling Detection: How would you detect and correct poor code styling 2. Security Vulnerabilities: How would you identify and fix security issues like SQL injection and XSS 3. Developer Interface: How would you design an interface for developers to review and approve AI suggestions 4. CICD Integration: How would you integrate this system into a CICD pipeline 3. Third Round: Can you explain your projects and let me know more about your skillset? Tell me more about your NLP & LLM based projects? Have you explored the latest large language models(LLMs) ? Have you fine tuned or retrained any LLMs? What LLMs have you worked with? Have you worked on any NLP related projects before LLM based on BERT or other models? Have you worked with RAG (Retrieval-Augmented Generation) pipeline? If yes, can you explain how? What is transformer architecture? Have you worked with Vector Databases? If yes, which ones have you worked with? What is hallucination in Al models? Why are you interested in Al ? How would you rate yourself in Deep Learning out of 10 ? Have you worked on model deployment on any cloud platform ? like GCP, AWS, Azure How would you rate yourself on Python out of 10? Have you worked on any Python framework? (Django, Flask, FastAPI) Preferable skills Do you have experience with full stack skillset (MERN stack)? Preferable skills What is the dependency on you right now? (Appicable to Full-timers) What is F1 score and ROC curve Find the longest increasing subsequence (LIS) in an array of integers. The longest increasing subsequence is the longest subsequence of a given sequence in which the subsequence's elements are in strictly increasing order. arr = [10, 9, 2, 5, 3, 7, 101, 18] The longest increasing subsequence is [3, 7, 101]
Projects, experience, papers.., coding question
Explain me loss? Go over attention for someone who doesn’t know the technical details.
Resume questions and ML questions
- 6 hard quiz - questions with 4-6 T/F options (not necessarily from your everyday ML algorithms). - 1 medium leetcode problem (from their list of most repeated questions) - 2 implementation questions. Mine were KNN and Kmeans. - One neural network hand-calculation (easy) Overall the assessment was thorough, but the time limit is unreasonably low! As someone with tons of coding and ML experience, I could say you had to be top 10% of Stanford graduates so that you can come close to getting all the questions right in 70 mins.
First Round: With one interviewer, I was asked a set of Python MCQs covering concepts like mutable vs immutable types, list comprehensions, and variable scope. Additionally, I solved DSA problems focused on arrays and strings to demonstrate my problem-solving skills and optimization techniques. Second Round: Again, with a single interviewer, I worked on machine learning coding problems and discussed core ML concepts, including algorithms, model evaluation, and practical ML applications.
Q. How will you explain overfitting and underfitting to a non-technical person?
asked to code backpropagation in numpy.
Viewing 2151 - 2160 interview questions