How would you add single sign-on to an enterprise AI assistant?
Deployment and LLMOps · Pro 1
Short answer
Integrate with the client's identity provider, such as Microsoft Entra ID or Okta, using OpenID Connect (or SAML if that is what they use). Users sign in with company accounts, the app receives an ID token with their identity and groups, and you use those groups to enforce document permissions in retrieval. Add provisioning so leavers lose access.
Deeper answer
OAuth 2.0 is for delegated access to APIs; OpenID Connect adds identity on top, which is what sign-in needs. Validate tokens properly: signature, issuer, audience and expiry.
Group membership can change, so refresh it on each session or sync it, and make sure the RAG permission filters use the same identity.
What a strong answer covers
- Integrates with the client's identity provider
- Uses OIDC or SAML
- Validates tokens and reads groups
- Groups drive retrieval permissions
- Provisioning and logging
Answer it out loud in the mock interview about 1.5 min
Follow-ups to expect
- What is the difference between OAuth 2.0 and OpenID Connect?
- How do you make sure a leaver loses access?
Common mistakes
- Building your own password system
- Not linking identity to retrieval permissions
Learn it
- When would you serve predictions in batch, and when online?
- How would you set up CI/CD for an LLM application?
- What is the difference between data parallelism and model parallelism?
- Why does experiment tracking matter, and what do you record?
- Your model scored well offline but performed badly after launch. What could explain it?
- Explain Kubernetes pods, deployments and services, and why you would use Kubernetes for AI apps.
- What is the KV cache, and why can it cause out-of-memory errors when serving an LLM?
- How would you load test an LLM feature and set a latency budget?
- Your model provider has an outage during peak hours. What should already be in place?
- How do you manage API keys and secrets across environments?
- How do you detect and handle model drift in production?
- A bank wants your AI system but says no data may leave its network. What are the options?
- Why does an LLM reply have a wait before the first word and then a different speed after it?
- What is quantisation, and what are the trade-offs?
- When would you use a semantic cache for an LLM app, and what can go wrong?
- How would you serve an LLM feature as an API for a web app?
- How would you serve a model that is too big for one GPU, and how would you scale it to more users?
- How do you version prompts and models so you can roll back safely?
- What would you capture in traces for an LLM application?
- Why use Docker for an AI application?
- What does an LLM gateway do, and when do you need one?
- Why use vLLM to serve an open-weight model instead of a simple Hugging Face script?