Machine Learning Engineer · Phone Screen · Technical Discussion

Meta — Machine Learning Engineer ✅ Passed

Level: Senior-Level

Round: Phone Screen · Type: Technical Discussion · Difficulty: 3/10 · Duration: 60 min · Interviewer: Friendly

Topics: Machine Learning, Large Language Models, Attention Mechanisms, Transformer Architecture, Positional Encoding, Dropout

Location: San Francisco Bay Area

Interview date: 2026-01-20

Summary

This phone screen focused on my experience with agent evaluation and LLM SFT/DPO, as well as technical questions on attention, transformers, positional encoding, and dropout.

Details

In the phone screen, I was asked to do a deep dive into my projects, especially regarding agent evaluation and LLM SFT/DPO. They asked about how I prepared and synthesized data, and what challenges I faced.

I was also asked to describe attention and transformer architecture, including attention weight calculation, masking, and the components within each transformer block. Other questions included comparing the pros and cons of different positional encoding methods (absolute, relative, RoPE), and the difference in dropout implementation between training and inference.