Career Roadmap
Generative AI Engineer: Zero to Hero
This roadmap guides you from AI foundations to professional generative AI engineering. You will build RAG pipelines, design agentic systems, deploy production AI workloads, and secure enterprise AI applications. The roadmap covers both Microsoft Azure (AI-901, AI-103, AI-200) and AWS (AIF-C01, AIP-C01) tracks. Critical 2026 updates: AI-900 and AI-102 retire June 30, replaced by AI-901 and AI-103. AZ-204 retires July 31, replaced by AI-200. Use ExamOS practice quizzes to track progress at every stage.
Who is this roadmap for?
This roadmap is designed for Software Engineers transitioning into AI who already write Python, as well as Cloud Engineers working with Azure or AWS who want to add AI specialization. Additionally, it is tailored for Data Engineers building AI pipelines, as well as ML Engineers expanding into generative AI and agentic systems who require hands-on credential validation.
Skills You'll Develop
Build AI applications with Python, async patterns, REST APIs, and SDK integrations.
Design and deploy AI solutions using Foundry hubs, agents, RAG, and evaluation frameworks.
Build GenAI applications with Bedrock Agents, Knowledge Bases, Guardrails, and evaluation.
Implement retrieval-augmented generation with Azure AI Search or OpenSearch Serverless.
Build multi-agent systems with Semantic Kernel, Bedrock Agents, and MCP integration.
Secure, monitor, and optimize AI workloads for cost, latency, and responsible AI compliance.
Target Roles in this Roadmap
- Generative AI Engineer: Builds and deploys LLM applications, RAG pipelines, and agentic systems
- AI Agent Developer: Designs multi-agent systems with Semantic Kernel or Bedrock Agents
- RAG Engineer: Specializes in retrieval-augmented generation and vector search
- AI Platform Engineer: Builds production AI infrastructure with containers, pipelines, and observability
- AI Security Engineer: Secures AI workloads with guardrails, prompt injection defense, and compliance
Typical Employer Categories
- Technology & Software Firms: Product AI features, internal AI platforms, developer tools
- Financial Services: Fraud detection, document processing, compliance automation
- Healthcare & Life Sciences: Clinical documentation, medical imaging, drug discovery
- Retail & E-Commerce: Personalization, inventory optimization, customer service agents
- Consulting & Professional Services: Building GenAI solutions for enterprise clients
- Cloud Providers & Partners: Azure Foundry and AWS Bedrock implementation specialists
The Certification Path
Recommended Path
| Cert | When | Why |
|---|---|---|
| Microsoft AI-901 (Azure AI Fundamentals) | Month 1-2 | AI and GenAI fundamentals. Optional if you already work with AI. |
| Microsoft AI-103 (Azure AI Apps and Agents Developer) | Month 4-8 | The core Azure AI credential. Validates Foundry, agents, RAG, and multimodal AI. |
| Microsoft AI-200 (Azure AI Cloud Developer) | Month 8-12 | Back-end depth. Covers containers, event-driven pipelines, vector data, and observability. |
| AWS AIP-C01 (GenAI Developer Professional) | Month 8-12 | AWS GenAI depth. Validates Bedrock Agents, Knowledge Bases, Guardrails, and evaluation. |
Optional Foundation
| Cert | When | Why |
|---|---|---|
| AWS AIF-C01 (AI Practitioner) | Month 1-2 | AWS AI fundamentals. Alternative to AI-901 for AWS-focused candidates. |
Milestones: Junior → Mid → Senior
| Level | Milestone | When |
|---|---|---|
| Entry Level | Python fluency + AI fundamentals + basic RAG understanding | Month 2-3 |
| Practitioner Level | AI-103 or AIP-C01 + production GenAI proficiency | Month 8-10 |
| Specialist Level | AI-103 + AI-200 or AIP-C01 + enterprise GenAI engineering skills | Month 12+ |
Step 1 - AI and engineering foundations
Build the programming and AI foundations that every GenAI engineering task depends on. Python fluency and AI fundamentals are non-negotiable.
~1-2 months~1-2 months
Step 1 - AI and engineering foundations
Build the programming and AI foundations that every GenAI engineering task depends on. Python fluency and AI fundamentals are non-negotiable.
- Python proficiency: functions, classes, async/await, virtual environments, SDK documentation, FastAPI basics
- REST API fundamentals: HTTP methods, authentication headers, request/response patterns, rate limiting
- JSON and data handling: parsing, serializing, working with nested structures, schema validation
- Git fundamentals: version control for AI projects, managing prompts and configuration securely
- AI concepts: LLMs, tokens, context windows, temperature, system prompts, embeddings, hallucination
- Responsible AI principles: fairness, reliability, privacy, inclusiveness, transparency, accountability
- Microsoft Foundry overview (Azure track): hubs, projects, model catalog, deployments
- Amazon Bedrock overview (AWS track): model access, knowledge bases, agents, guardrails
💡 AI-901 is the Azure entry credential. It replaced AI-900 on June 30, 2026 and is built around Foundry rather than individual services. 40-60 questions, 45 minutes, 700/1000 passing score.
💡 AIF-C01 is the AWS equivalent. 85 questions, 90 minutes, 700/1000 passing score, $100 USD.
🏁 Entry Level Checkpoint: You can write Python code, call LLM APIs, and understand core GenAI concepts.
🛠 Project Ideas
- ▸Write a Python script that calls an LLM API with a system prompt and returns structured JSON output.
- ▸Create a simple REST API using FastAPI that accepts a prompt and returns an AI-generated response.
Step 2 - Core GenAI engineering (choose your platform)
Build professional GenAI engineering skills on your primary platform. This step covers the full AI lifecycle—RAG, agents, evaluation, and production deployment.
~4-6 months~4-6 months
Step 2 - Core GenAI engineering (choose your platform)
Build professional GenAI engineering skills on your primary platform. This step covers the full AI lifecycle—RAG, agents, evaluation, and production deployment.
- Microsoft Azure track (AI-103): Foundry Agent Service, Semantic Kernel, multi-agent architectures, Entra Agent ID, RAG with Azure AI Search, Document Intelligence, multimodal (Vision, Language, Speech), Content Safety, prompt injection defense, cost and latency optimization
- AWS track (AIP-C01): Bedrock Agents, action groups, Knowledge Bases (OpenSearch Serverless), Guardrails (content filters, PII, grounding), model evaluation (automatic and human), MCP integration, production GenAI architecture
Certifications
💡 AI-103 replaced AI-102 on June 30, 2026. Generative AI and Agentic Solutions is 35-40% of the exam—the primary domain. Build real agents and RAG pipelines before your exam.
💡 AIP-C01 is described as the hardest AWS AI certification. 65 scored questions, 130 minutes, 750/1000 passing score, $300 USD. Requires genuine hands-on Bedrock experience.
🏁 Practitioner Level Checkpoint: You have passed AI-103 or AIP-C01. You can design, build, and deploy production GenAI applications.
Step 3 - Production AI engineering (back-end depth)
Build the back-end cloud infrastructure that AI applications run on—containers, event-driven pipelines, vector data services, security, and observability.
~3-4 months~3-4 months
Step 3 - Production AI engineering (back-end depth)
Build the back-end cloud infrastructure that AI applications run on—containers, event-driven pipelines, vector data services, security, and observability.
- Azure track (AI-200): Containerized AI workloads (ACR, ACA, AKS), event-driven pipelines (Event Grid, Service Bus, Functions), vector databases (Cosmos DB, PostgreSQL pgvector, Redis), API Management for AI (rate limiting, token quotas), distributed observability (Application Insights, OpenTelemetry)
- AWS track (optional): Building production Bedrock pipelines with Lambda, Step Functions, and CloudWatch; vector search with OpenSearch Serverless
Certifications
💡 AI-200 replaced AZ-204 on July 31, 2026. It covers Azure AI and Cloud Services (25%), Data Services for AI (20%), Event-Driven Pipelines (20%), Secure and Scalable Deployment (15%), and Observability (20%).
🏁 Specialist Level Checkpoint: You have passed AI-200. You can design and deploy production AI infrastructure on Azure.
Step 4 - Exam consolidation, security, and career advancement
Consolidate your skills, validate with your target certifications, and plan your next career move. Security, responsible AI, and emerging patterns like MCP are the differentiators.
~1 month~1 month
Step 4 - Exam consolidation, security, and career advancement
Consolidate your skills, validate with your target certifications, and plan your next career move. Security, responsible AI, and emerging patterns like MCP are the differentiators.
- AI security: prompt injection defense (direct and indirect), content safety at scale, input/output validation
- Cost management: token budgeting, prompt compression, model routing, semantic caching
- Responsible AI: bias testing, red-teaming, model cards, EU AI Act compliance
- Emerging patterns: MCP (Model Context Protocol), computer use agents, long-horizon task completion
- Follow-on paths: Microsoft SC-500 (AI Security), AWS AI Security, CKA (for AI infrastructure)
💡 Prompt injection is the OWASP
💡 Consistent performance above 80% on Legend mode across five consecutive ExamOS sessions is the clearest readiness signal.
💡 MCP (Model Context Protocol) is emerging as the standard for agent-tool communication. Familiarity with it future-proofs your agent architecture knowledge.
🏁 Specialist Level Checkpoint: You have passed your target certifications. You can design, build, and secure enterprise-grade GenAI solutions.
🛠 Project Ideas
- ▸Build a complete production GenAI application with security controls, cost monitoring, and evaluation framework.
Final Step - What this path actually builds
This roadmap moves you from AI foundations to production-grade generative AI engineering. AI-901 or AIF-C01 establishes your AI vocabulary. AI-103 or AIP-C01 validates your ability to build GenAI applications with RAG, agents, and evaluation. AI-200 adds the back-end infrastructure depth that makes AI applications production-ready. The 2026 Microsoft transitions—AI-900 to AI-901, AI-102 to AI-103, AZ-204 to AI-200—reflect the shift to Foundry-centered, agent-focused, and infrastructure-aware AI engineering. The most important skill is not knowing any single service, but understanding the architecture patterns that work across platforms. Build real agents and RAG pipelines. Measure your readiness with ExamOS. Book when your reasoning is sharp and your architectures are defensible.
Final Step - What this path actually builds
This roadmap moves you from AI foundations to production-grade generative AI engineering. AI-901 or AIF-C01 establishes your AI vocabulary. AI-103 or AIP-C01 validates your ability to build GenAI applications with RAG, agents, and evaluation. AI-200 adds the back-end infrastructure depth that makes AI applications production-ready. The 2026 Microsoft transitions—AI-900 to AI-901, AI-102 to AI-103, AZ-204 to AI-200—reflect the shift to Foundry-centered, agent-focused, and infrastructure-aware AI engineering. The most important skill is not knowing any single service, but understanding the architecture patterns that work across platforms. Build real agents and RAG pipelines. Measure your readiness with ExamOS. Book when your reasoning is sharp and your architectures are defensible.
Certifications
Final Thoughts
💡 Total: 6-12 months at 2 hours/day (4-8 months at 3-4 hours/day)
💡 AI-901/AIF-C01: 2-3 weeks for candidates with AI exposure (optional).
💡 AI-103/AIP-C01: 4-6 months. The core GenAI credential—invest the most time here.
💡 AI-200: 3-4 months. Requires genuine back-end development experience.
💡 Readiness: Consistent 80%+ on Legend mode across five sessions is your signal to book.
Honest Timeline
| Path | Minimum Study Pace | More Realistic Pace |
|---|---|---|
| AI-901 only (optional) | ~1 month | ~2 months |
| AI-103 only (Azure track) | ~4 months | ~6 months |
| AIP-C01 only (AWS track) | ~4 months | ~6 months |
| AI-103 + AI-200 (Azure full track) | ~6 months | ~10 months |
| AI-103 + AIP-C01 (multi-cloud) | ~8 months | ~12 months |