Together AI applied ML systems coding interview

Round 5 of 7

See the whole loop
Length
60 min
Format
Live coding
They grade
Working model-serving code and clear performance tradeoffs; research depth is not needed
AI use
One guide says some rounds allow or encourage Cursor or Copilot and then grade how you check the AI's code
Heads-up
Implement part of an inference engine: an attention primitive, streaming generation, request batching or KV cache management.

Questions

No Together AI reports yet. Other labs ask: