Google 2026 New Grad SWE AI/ML Coding and ML Domain Interviews

Role: Software Engineer, PhD Early Career — AI/ML
Result: Advanced to Team Match
Report time: February 2026
ML specialization: Applied ML

Interview process

1. Recruiter call on October 13.
2. Selected the ML domain on October 16.
3. Completed R1 coding and behavioral interviews on November 24.
4. Passed R1 on December 8.
5. Completed the California onsite on February 5.
6. Received positive feedback and entered Team Match on February 25.

R1 coding

Given a collection of log files, return the ten IP addresses that appear most frequently.

The candidate used a frequency map followed by a Top-K structure.

Follow-ups asked:

1. What changes when the input file is very large?
2. What if the data cannot fit in memory?
3. How would the solution work in a distributed environment?

The interviewer spent additional time on these follow-ups because the candidate had prior distributed-computing and storage experience.

Practice: Top-K IP Addresses from File Records

Onsite coding

The coding problem was a Basic Calculator II variation supporting addition, subtraction, multiplication, and division, with additional behavior for empty input.

The interviewer expected the candidate to:

• Explain the complete approach on a whiteboard before coding.
• Confirm edge cases proactively.
• State the time and space complexity clearly.
• Prefer a complete, logically consistent implementation over rushing.

Practice: Basic Calculator II With Empty Input

ML domain interview

The ML round focused on fundamentals and applied understanding:

• Bias–variance tradeoff
• Logistic regression
• Gradient descent
• Cross-entropy
• L1 and L2 regularization
• Forward and backward propagation

The follow-ups went beyond definitions. For L1 regularization, the candidate had to explain why it creates sparse parameters and draw the constraint geometry. For backpropagation, the candidate had to explain the loss, chain-rule gradient flow, and parameter updates.

The candidate described the round as a test of depth and communication rather than memorized definitions.