Q: What are the state-of-the-art in object recognition and human pose prediction areas? Q: How can we fix the corner cases of our CNN/other computer vision models Q: What is your idea how to improve accuracy close to 100% Q: Why Woven Planet? Why Japan? Do you know Japanese? Other questions about state-of-the-art Background-related questions...
Basic ML things such as loading images from disk, normalizing them, passing them through a pre-trained model, computing embeddings etc.
BERT, OCR, NAtural Language Processing, Deeping Learning.
There were no questions being asked.
Interview questions are fine. But the unprofessionalism left a strong impression on me.
If there were a buh, how would you deal with them
Bagging vs Boosting.Bias Variance tradeoff.
R1: It was a python test. they had given me table in consisting the information about all tests given by students. each entry in data was representing the test score. so I had to group them using pandas and pivot them by subject wise. I wrote the solution anyhow. then interview was based on basic of python and my solution. Round2: After clearing the round1 i got invitation for the r2 i.e NLP round. this time they had given me the dataset of customer problems and sub problems. I had to train a NLP model that can classify the type of problems into different category. I was learning NLP at that stage so I could not perform well in R2. but yeah I learned alot in these process.
Working of Autoencoders, VAEs, GANs etc.
Questions regarding specific machine learning models that I have never heard of before
Viewing 5601 - 5610 interview questions