Ho presentato la mia candidatura tramite un selezionatore. La procedura ha richiesto 2 settimane. Ho sostenuto un colloquio presso KnowDis Data Science (Calcutta) nel mese di apr 2025
Colloquio
There were 3 Technical Rounds, 1st was one Python and 2nd and 3rd are complete depth of Deep Learning.
Python round covered coding K-Mean algorithm from scratch and a Leetcode medium level question on DFS.
Second round they asked questions around RNN, LSTM, Neural Network Initialization and Optimization
Second round they covered BERT my project in depth
Domande di colloquio [4]
Domanda 1
Code K-Mean algorithm. They gave dummy data of 5 points, and two centers. I had to code and provide the output cluster centers.
Issues of RNN?
How can you solve them without moving to LSTM?
Gates of LSTM?
All the components of Transformer? What is the use of positional embedding? What is Multi-head attention? etc.
Different type of NN initializers, why they work?
Best optimizer? How they work?
All the components of Transformer? How are they trained
What is encoder vs decoder architecture?
What is masked decoder?
What is auto regressive LLMs?
What is the architecture of Distill BERT?
