A production-grounded Ollama interview-questions guide — Modelfile DSL, GGUF quantization choice, concurrency knobs, the 2024-2026 CVE chain, and the vLLM crossover, with cited postmortem sources.
A practitioner-shaped LangGraph interview guide built around the 2024-2026 production stack — StateGraph internals, supervisor-vs-swarm patterns, checkpointer choice, and human-in-the-loop interrupts.
DSPy interview questions calibrated for 2026 loops at Databricks, JetBlue, Mistral, and Stanford-adjacent labs — covering MIPROv2 internals, the declarative-vs-imperative wedge with LangChain, and production failure modes from real practitioner postmortems.
Real OpenAI Codex CLI interview prep grounded in the April 16, 2025 launch (4M weekly active users by April 2026 per Sam Altman); explicit disambiguation from the deprecated 2021 Codex API; planning patterns (Plan mode, Reverse Interview, AGENTS.md); approval modes (–full-auto blast radius); GPT-5.5 reasoning levels; Codex vs Claude Code architecture comparison (Rust + GPT-5.5 vs TypeScript + Claude Sonnet 4.x); enterprise governance + MCP centralization; scenario-based postmortem questions for senior interviews.
Real Claude Code interview prep for hiring software engineers: Anthropic Feb 24 2025 CLI launch, CLAUDE.md project memory, slash commands (/agents /skills /hooks), Task tool sub-agents, MCP integration, permissions model + safe-defaults, plugins + skills extension points, 2026 features (worktrees, channels, checkpoints, SDK), prompt-injection research, Anthropic agentic-misuse paper, and scenario-based behavioral questions (what would you do if Claude deleted a file in production?). Distinct from the SFEIR Institute 27-Q list — covers 2026 features SFEIR lacks plus the named-incident postmortem layer senior interviewers probe.
Real MCP (Model Context Protocol) interview prep grounded in the November 2024 Anthropic launch, protocolVersion 2024-11-05 JSON-RPC 2.0 spec, the three transports (stdio, SSE deprecated, Streamable HTTP current), server primitives (tools/resources/prompts), client primitives (sampling/roots), the OpenAI March 2025 and Google DeepMind April 2025 adoption timeline, AWS AgentCore Gateway integration, and the Wiz security briefing on tool-poisoning + rug-pull + Asana MCP cross-tenant leak postmortems.
Real AWS Bedrock AgentCore interview prep grounded in named-incident postmortems: October 2025 us-east-1 DynamoDB DNS race condition cascading outage that took down Lambda + Bedrock, Asana MCP cross-tenant data leak May 2025, Palo Alto Unit42 ‘Cracks in the Bedrock’ DNS-based sandbox escape April 7 2026, Unit42 IAM role-chaining God Mode April 6 2026, and AgentCore silent session restart November 2025. Covers all 10 AgentCore components (Runtime, Memory, Identity, Browser, Code Interpreter, Gateway, Observability, Evaluations, Registry, Policy/Cedar) and the AgentCore Runtime vs Lambda architectural distinction (stateful, longer execution, Docker-packaged) senior interviewers actually probe.
Real AI engineer interview prep grounded in production specifics: Air Canada chatbot tribunal ruling (Moffatt v. Air Canada, February 2024, $812.02 CAD), Mata v. Avianca ChatGPT hallucination sanctions (June 22, 2023), Samsung leak, DPD chatbot, Stanford 58-88% hallucination study, Anthropic multi-agent research (90.2% / 15x / 80% task-to-task variance), MCP risk taxonomy, and 2024-2026 hiring shifts at Anthropic, OpenAI, Cohere, and Hugging Face.
Real practitioner-reported LangChain interview questions for 2026 — anchored to the CVE-2025-68664 deserialization exploit, Octomind’s public abandonment, the legaltech LCEL swap, and the ~15-call tool-stack degradation signal that interviewers actually test.