← Back to Directory

Reflection AI

Small

Screna has 2 Reflection AI interview write-ups. The most reported round is the technical phone screen (2).

2 interview notes · updated Jul 2026

Reflection AI·Software Engineer·Technical Phone Screen

Jun 2026
Research Engineer interview at ReflectionAI covering LLM fundamentals pretty deeply, transformer internals, RL-based post-training methods, and then a specific technical drill-down on parallelism strategies for inference.
  • If you have a single decoding request to serve (no batching), which parallelism strategy would you pick among data parallelism, tensor parallelism, and pipeline parallelism, and why? Walk through the tradeoffs in latency, memory, and underutilization.

“This is the kind of question that sounds easy until you start talking and realize you've been thinking about parallelism mostly in the training context.”

View Post

Reflection AI·Software Engineer·Technical Phone Screen

May 2026
Coding question for a Research Engineer role at ReflectionAI. The problem was a sequence transformation that sounds deceptively simple but pushes you toward thinking in a vectorized/tensor style rather than just looping through.
  • Given a binary 0/1 sequence, transform it so every 0 becomes -1 and every 1 is replaced by the index of the contiguous block of 1s it belongs to (blocks numbered left to right starting from 0). For example, `0010110111` becomes `-1,-1,0,-1,1,1,-1,2,2,2`. Implement this, preferably in a vectorized or tensor-style approach.

“My first instinct was a plain loop with a counter and I coded that up pretty fast, but the nudge toward vectorized solutions was clearly the real ask.”

View Post