Phone round:

• Implement rand07 using rand01
• Implement unbiased rand01 from biased rand01
• Implement randxy (random number in the range x - y inclusive)
• ML basics: logistic regression, neural nets, gradient descent etc.

Onsite round 1:

• Project discussion/ Ph.D. research
• ML basics such as neural nets, gradient descent etc.

Onsite round 2:

• Round with manager. Past projects + implement a job reccomendation system. Also, asked questions related to scaling systems.

Onsite round 3:

• Convert a roman numeral to integer
  They also said they will ask integer to roman. However after I solved first they said what if the input number roman is not valid? They didn't want me to solve the other question since I started having trouble solving all the edges cases. I wasn't able it completely solve it, i think it is very tricky. For example, here are some invalid cases: XXL, IL, LL etc.

Onsite round 4:

• What is batch normalization?
• What is the loss function for SVM?
• Given a sorted array, apply the transformation ax^2 + bx + c and return new sorted array. Follow up: how do we extend the idea to cubic polynomial?

Onsite round 5:

• Desing a system to recommend music artists to followers. Suppose we have an artist that doesn't have enough followers (eg. 100) and he wants to go for a tour. How would we use our limited budget to publicize this artist to users (eg. 1M)?