Bro. Who ever asks for DSA for a MLE or Data Scientist is not good working for. So chill. I write to them that 'this is a very bad interview practise and reject the interview' (it's waste of time for me).
Besides look, everytime I try to switch and go for new role, I also get rejected initially, it's about you getting acclaimed to the hiring environment. But after each tranche of failures, I have managed to grab 50-100% jumps. So getting rejected is part and parcel. It will prepare you better for next one. That fear is good. So accept it and let it guide you to prepare better. It's alright
Now coming to actual stuff other than these gyaans everyone has given--
- Make a list of basics to prepare for and each day cover those topics.. I will just jot down a comprehensive view so that it will help both me and you
-linear regression - assumptions
-pValue significance (statquest videos)
-heteroScedasiticy shyt
-lasso vs ridge, why lasso reduces ur features (get this down well)
-AB testing (for every project u have worked on you need to tell how you designed ur AB experiment)
Learn power analysis
How to get minimum sample size-- statquest again
-bias variance tradeoff
-confidence intervals, standard error
-Rocr vs PR curves
-central limit theorem, z test, t test, 2tailed test bla bla
-PCA (though no one uses it, some interviewers are shite)
-regularization
- why use sigmoid in logistic?
Loss functions- MSE ve MAE, vs Mape ADV and disadv of each
Sigmoid graph
Decision Trees, entropy, info gain, gini
XGBoost, Random forest
Isololation forests
Get the xgboost loss function and how boosting exactly works,
Regularisation in xgboost
Be prepared to code GBM as well. It's simple
Neural nets
Why do you need activation functions?
Gradient descent, derivation as well
Gradient descent through time
Cross entropy loss and it's derivation(read max likelihood function)
Auto encoders- read how it helps anam