J'ai postulé en ligne. J'ai passé un entretien chez Papercup.AI
Entretien
The interview was held on Zoom, with technical questions about Deep Generative Networks. The interviewer was very nice but they wanted to hear key words even though the questions were very vague.
Questions d'entretien [1]
Question 1
What is a DGN, tell me your experience on GANs and VAEs.
J'ai postulé via un établissement d'enseignement supérieur ou universitaire. Le processus a pris 1 jour. J'ai passé un entretien chez Papercup.AI (Londres, Angleterre) en janv. 2022
Entretien
First talked about previous Machine learning projects, anything could be fine, there were no specific titles. Then talked about the motivation and why I wanted to work in this company.
J'ai postulé en ligne. J'ai passé un entretien chez Papercup.AI (Londres, Angleterre) en mai 2021
Entretien
1. motivational call with HR + background check (~30mins)
2. introductory interview with 2 ML engineers (~ 1hrs. Talking about personal projects much more in specifics than with HR. Additional questions on VAEs and Attention mechanism. Open-ended questions on how to use VAEs and Attention to tackle speech translation.
3. Coding interview, re-implementing Glow paper on MNIST dataset in Jupyter Notebook and Pytorch (~45mins).
4. Take home papers to study (1: Neural Processes, 2: VAEs for speech generation). 20 min presentation on one of the 2 papers + 30 min questions on both papers.
Questions d'entretien [1]
Question 1
Describe a VAE in detail
Describe Attention and Seq-2-Seq modelling in detail
Describe how you would combine the two above to obtain a text to speech generation pipeline
Describe Neural Processes paper in detail. Questions about Stochastic Processes and benefit of moving to neural processes. ELBO derivations/summary.
How to align text sequence to sequence of phonemes (different lengths)