1)Technical screening
Deep-dive interview on the candidate’s claimed stack. LangChain specialists get scenarios covering chain composition, memory management, and agent design. OpenAI-native builders get token optimization, function calling, and structured output tests. Voice AI candidates get latency and turn-taking questions. Senior engineers (not recruiters) lead this stage.
2)Live pair programming
The candidate solves a chatbot-relevant problem in real time — typically a RAG retrieval issue, a prompt template bug, or an agent orchestration failure. We grade for how they reason through ambiguity, not whether they memorized syntax.
