How Runway Interviews in 2026: Process, Sample Questions & What They Score

Background, motivation for Runway and generative AI, role fit, logistics. Several prep sites describe this the same way. TechPrep and Extern give the 30 minute figure; the other pages give no length.

Only DesignGurus lists this, as a chat some candidates report that covers role fit, past work and expectations. No other source mentions it and no source says what decides whether it happens, so the length and who gets it are unknown.

single source

Live coding in Python or TypeScript; TechPrep names CoderPad. TechPrep says clean, production-quality code matters; DesignGurus says working code with clear reasoning. JobMentis describes LeetCode-medium problems under time pressure. TechPrep also says candidates in 2026 noted AI assistants such as Cursor or Copilot may be permitted. ML candidates may get ML coding instead: one Aug 2025 ML Engineer candidate had two 45-minute phone screens, ML coding then ML system design.

TechPrep lists two coding rounds: algorithms and data structures at medium to hard level, plus object-oriented design and data modeling. TechInterview also lists two onsite coding rounds without detail. JobMentis lists a single onsite coding round (LeetCode-hard, debugging). DesignGurus says some candidates report a code review round where they read and improve existing code. Extern (an internship guide with estimated dates and no cited sources) lists pair programming and code review and estimates half a day to a full day for the whole onsite. No source gives per-round lengths. DesignGurus says round order and count vary by role and by team.

Domain-specific prompts: video generation pipelines, GPU resource management, credit billing, large media files. ML candidates may get ML system design instead. The one hour figure comes from one DesignGurus article only. Placement differs by source: JobMentis lists system design as its third round, before its onsite coding round.

TechPrep describes a presentation or deep technical discussion of a complex project you built, where interviewers probe architectural decisions, tradeoffs and how you handled scale or complexity. DesignGurus frames it as the hardest problem you solved and the choices you made. TechInterview names the round without detail. No source gives a length.

Reported focus: ownership, creative taste and customer empathy for creative professionals (TechPrep, TechInterview), plus collaboration, feedback and pace (DesignGurus). JobMentis lists ownership, influence and conflict resolution. No source gives a length.

Research candidates reportedly discuss a paper. Extern adds deep learning fundamentals and a PyTorch coding screen, and places these in the technical assessment; DesignGurus and TechInterview do not say where the paper discussion sits in the loop. No source gives a length.

Only Extern lists this: conversations with senior leadership or co-founders about culture, vision and research direction, estimated at 30 to 60 minutes. That guide covers an internship cycle Runway does not formally advertise, calls its dates informed estimates, labels its calendar funnel as expected, and cites no source for this stage. No other source mentions it. Treat it as unconfirmed.

single source

Sources disagree: about 3 weeks (DesignGurus), 2 to 4 weeks (Extern, which cites a Glassdoor average of 20 days but gives 80 interviews in one place and 97 in another), and 4 to 5 weeks (TechPrep, TechInterview). DesignGurus says the loop is about five to six conversations; the TechPrep and TechInterview stage lists add up to seven (recruiter screen, coding screen, five onsite sessions).