Zycus Interview Experience | AI/Machine Learning Engineer

Round 1: Assignment.  

You will get an assignment form HR to solve. For my case it was extraction of data from doc format. The crux of clearing this round lies in labeling the data after you have extracted it and then training a NLP model to get desired fields. They are just concerned about your approach, even you don't complete the 100% of assignment it's okay. 

Round 2: Webex based video conferencing. 

My interview was of 1.5 hours (3 rounds) . They asked majorly NLP sections, what are embeddings, skipgrams and cbow, how will you detect outliers in case of supervised and unsupervised learning. What type of model will you use given a case study. 

They will also check basic machine learning concepts and then they will start to dig deep. Be ready as they will open your project and will ask why did you write this particular line of code only, why not something else. 

Whatever the project you have mentioned they will understand it and frame some more questions on that. Like they asked me why have I used Ridge regression why not lasso or any other method. How does the word embedding map into the latent space, Glove embedding details. What is difference between Glove embedding and word2vec. Which will you prefer and why. 

Round 3:  In this main focus on how did I implement my project. They gave me case scenario as suppose you are a ML head of  Apollo hospitals where you are given 4 independent features such as ECG, systolic rate, disytolic rate, blood flow (could be any four independent parameter) and given upon this you have to predict whether or not the person will have a cardiac arrest. How will you approach this problem. Sequence modelling question. Then further they made modifications in the question that if certain details upto 60 timesteps is known then how, like they follow the pattern of cardiac arrest how will be sure that at timestep 120 the person will suffer a cardiac arrest. Rest they asked questions on Squential modelling. 

Round 4: Last they will ask about your assignment and data extraction, different ways, how will you extract table data from OCR. Focus on spacy models.