It has been 3 years working in Singapore since I graduated. During the period in Shopee, I learned a lot not only the coding skills but also the understanding of projects. Since Shopee has decided to shift the engineering center to China mainland, the projects that we Singapore team could do are becoming limited, which result in the thoughts to pursue other opportunities.
There are not many opportunities avaliable in Singapore as expected, some local big company like Lazada and Grab freeze their head count, Amreican company like Google and facebook also don’t have enough engineer head count. I got my permanent resident before I subimitted my CV, which helps me a lot to have more interview opportunities. Here are recent interview experience, unluckily I don’t have any offer due to short of time to fully prepare.
Interview
Tiktok
-
Machine learning engineer(recommendation)
Because of lack of recomendation-related experience, the interviewer in both 2 rounds doesn’t have too much interests on my experience. So the coding problem appears to be very important，but I didn’t prepare for a long time so coding performance sucks, failed the interview.
1st round: Leetcode 3(Longest Substring Without Repeating Characters)
2nd round: Leetcode 297(Serialize and Deserialize Binary Tree) -
Machine learning engineer(platform governance)
Bert: how to pretrain; Q, K, V
batch norm and layer norm
How does GBDT do classfication?
How to pretrain your own Bert?
Indeed
The coding problem is easy, merge 2 sorted array. Here are some machine learning related problems.
- Unfair coin(Bayesian probability)
- Overfitting and underfitting
- Variance, boosting and bagging
- Dropout mechanisum
Apple
Coding: Top k frequent elements
word2vec, Search System introduction, Bilstm-CRF
Huawei
leetcode: lognest valid parentheses, coin change
Grab
leetcode 1011
Reflection
To be honest, I am frustrated for a while, but I think I deserve it because I didn’t fully prepare the coding and related machine learning knowledge on the CV. Right now I don’t have chance and have to change my mind to look for new roles at the end of the year. It is not too bad because I have 4 months to prepare, this time I don’t have any excuse. First I need to know which position I should apply for, I think machine learning engineer(search or NLP) should be more related, it is also okay if the job needs me to still work on map-related projects except assignment field.
Preparation
Coding
Machine learning & Deep learning
- Underfitting and Overfitting
link - Bias and Variance
link - Generative Model vs Discriminative Model
link - Bagging vs Boost
link - Feature Selection
link - Bayesian Model
link - Logistic Regression
link - GBDT VS Xgboost VS LightGBM
link -
Metrics: (MSE,MAE, Logloss, AUC curve)
link - L1, L2 regularization
link - Hidden Markov Model
link - Conditional Random Field
link - Activate function
link - Optimization algorithm
link - Batch Norm VS Layer Norm
link
NLP
- RNN
link - Gradient vanish & gradient explode
- LSTM
Understanding LSTM Networks
LSTM模型与前向反向传播算法
Solving the Vanishing Gradient Problem with LSTM - Word2vec
-
Bert
1) The number of Bert parameters
2) Why does self-attention divide root of dk
3) [The comparison of all attentions] - Bi-lstm + CRF
CRF Layer on the Top of BiLSTM
Pytorch Code