Scale AI 2026 New Grad SWE Two-Day Onsite

The onsite ran across two days. There were two behavioral rounds and three technical rounds.

One round was Stateful Neuron Matrix Update. Another round was debugging an existing codebase. The backend practical asked for functions that called a model API, formatted the response, and validated the result. Python or TypeScript could be used for the debugging and backend rounds.

Their earlier phone screen used task scheduling and topological sorting. FastPrep has Task Processor: Dependencies and Deadlines and Format and Validate Model Response for practice.

They were rejected after the onsite.