Cloud, deployment & production
Use managed AI platforms (Bedrock / Vertex / Azure AI Foundry)
Provision models, manage IAM, use platform features (knowledge bases, guardrails, agents) on one hyperscaler.
~12 focused hours·intermediate
Tools: Amazon Bedrock, Google Vertex AI, Azure AI Foundry, IAM roles/policies, guardrails/knowledge bases
Market relevance — share of job ads asking for this
What employers mean
You should be able to…
- Provision access to a foundation model on Bedrock/Vertex/Azure AI Foundry with least-privilege IAM
- Set up a managed knowledge base/RAG feature on the platform instead of hand-rolling one
- Configure guardrails (content filters, PII redaction) on a managed AI platform
- Monitor token usage and cost per model/project on the platform's console
- Compare model options on one hyperscaler (latency, cost, context window) for a given use case
- Set up an agent or tool-calling flow using the platform's native agent framework
Needs first: Integrate LLM APIs into an application
Learn — free, link-checked
The few resources that matter
Read · beginner · 25 min · docs.aws.amazon.com
What is Amazon Bedrock?
Official intro to provisioning foundation models, knowledge bases and guardrails on AWS's managed AI platform. — AWS
Read · beginner · 25 min · cloud.google.com
Introduction to Vertex AI
Covers Vertex AI's model garden, pipelines and IAM setup — the GCP-equivalent managed platform employers reference in JDs. — Google Cloud
Read · beginner · 25 min · learn.microsoft.com
What is Azure AI Foundry?
Explains Azure's unified model catalog, agents and guardrails — the third hyperscaler platform Indian job posts name. — Microsoft Learn
Read · beginner · 25 min · docs.aws.amazon.com
What is IAM?
Least-privilege IAM is the single most-tested security concept in Indian AI interviews; the canonical explanation of roles vs policies. — AWS
Course · beginner · 90 min · cloudskillsboost.google
Google Cloud Skills Boost — Learning Paths
Free guided, hands-on labs (not just docs) for GCP and Vertex AI, with quizzes and a shareable badge for your portfolio. — Google Cloud
Read · intermediate · 20 min · cloud.google.com
GKE Quickstart
Get a real managed cluster running in minutes — useful for GPU node pool scheduling once you've learned the basics. — Google Cloud
Read · intermediate · 30 min · cloud.google.com
Deploy a Python service to Cloud Run
The fastest free path from a Python API to a public HTTPS endpoint with autoscaling — the most common 'ship it' step in Indian AI job tests. — Google Cloud
Read · intermediate · 30 min · learn.microsoft.com
Deploy your first container app
The Azure equivalent deploy path — useful since several Indian enterprise/BFSI employers standardize on Azure. — Microsoft Learn
Practice
Managed RAG Assistant on One Hyperscaler
Pick one platform (Bedrock, Vertex AI, or Azure AI Foundry) and build a small assistant that answers questions over a set of Hindi/English NBFC or insurance policy PDFs using the platform's managed knowledge-base/RAG feature, with a least-privilege IAM role and a guardrail (PII redaction or content filter) enabled. Document cost per 1,000 queries.
Done when
- Assistant answers correctly from the uploaded documents via the platform's managed RAG feature
- IAM role/service account used has only the permissions needed to invoke that model and knowledge base, nothing broader
- A guardrail (PII redaction or content filter) is enabled and demonstrated with a test query it blocks/redacts
- A short cost note estimating $/1000 queries based on the platform's published pricing
Prove it
Evidence a recruiter can check
- Screen recording or screenshots of the working assistant plus the IAM policy JSON used
- Architecture note explaining which managed features were used vs custom code
- Cost estimate with the platform pricing page linked as the source
Interview
Questions you'll get asked
- How would you set up least-privilege IAM for a service that only needs to invoke one Bedrock model?
- What's the difference between Bedrock Knowledge Bases and rolling your own RAG pipeline?
- How do you monitor and cap token spend per team/project on a managed AI platform?
- When would you choose a managed agent framework (Bedrock Agents/Vertex Agent Builder) over building your own orchestration?
- How do you configure content-safety guardrails on a managed platform, and what do they catch?
- Walk me through comparing two models on the same platform for cost vs latency trade-offs.