Skip to content
PREP FOR A SPECIFIC LOOP

Interview prep by company

Each page says what the company's own postings and published guidance support about its loop, with a confidence level, the posted pay where there is one, and the questions that map to it.

35 companies·7 with a published process·reviewed October 2026

Frontier model labs

8Inference and model serving teams for the labs' own models: fleet design, engines and implementation rounds.
OpenAIAnthropicGoogle DeepMindMetaxAI (SpaceXSI)Mistral AICohereSarvam AI

Inference providers and AI platforms

10Serving open models for customers at scale: engines, scheduling, disaggregation and debugging.
Fireworks AITogether AIBasetenModalAnyscalePerplexityCerebrasSambaNova SystemsDatabricksRed Hat

GPU and accelerator companies

8Kernels, runtimes and inference software for their own hardware: the most kernel-heavy loops.
NVIDIAAMDTenstorrentd-MatrixEtchedQualcommGoogle (TPU / Cloud AI infrastructure)AWS Annapurna Labs (Neuron)

Big tech

5Serving inside large product organizations, often with compilers and performance analysis in the loop.
GoogleMicrosoft (Microsoft AI)Amazon (AGI)LinkedInSnowflake

Trading firms

4Models on the critical path: latency, GPU performance and whole-system optimization.
Jane StreetHHudson River TradingTwo SigmaOptiver