Prerequisites Basic computer literacy and a willingness to document experiments.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Tokens, context windows and message roles Sampling, determinism and model limits Structured input/output contracts Latency, cost and privacy boundaries Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools Official model documentation Tokenizer viewer Python
Hands-on exercise Estimate token budgets for a set of realistic user requests and inspect truncation behaviour. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project A small model interface wrapper with typed request/response structures. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria The application handles context limits and records cost/latency safely. Evidence must be reproducible and tied to Generative AI.
Next: Prompt design & structured generation
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Instruction hierarchy and prompt templates Few-shot examples and constrained output JSON/schema validation Prompt injection awareness Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools Model API or local model JSON Schema Prompt test cases
Hands-on exercise Build a structured extraction task and test malformed, adversarial and ambiguous input. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project A prompt catalogue with schema validation and a small regression set. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria Invalid output is rejected or repaired without trusting raw model text. Evidence must be reproducible and tied to Generative AI.
Next: Retrieval, tools & context augmentation
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Retrieval interfaces and source attribution Tool/function calling contracts Permission boundaries for tools Context assembly and caching Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools Vector store where applicable Tool schemas API client
Hands-on exercise Create one read-only tool and one grounded retrieval flow with explicit access controls. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project A model-assisted workflow that cites its data sources and limits tool scope. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria Tool inputs are validated and every factual response can be traced. Evidence must be reproducible and tied to Generative AI.
Next: Evaluation, guardrails & red teaming
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Golden datasets and rubric evaluation Hallucination and citation checks Prompt injection and data exfiltration risks Human escalation Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools Evaluation harness JSON schema validator Test fixtures
Hands-on exercise Create a small evaluation set and run adversarial prompts against the workflow. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project An evaluation report with pass/fail thresholds and known limitations. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria Quality and safety criteria are measurable and regressions are repeatable. Evidence must be reproducible and tied to Generative AI.
Next: Application orchestration & state
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Workflow graphs and state transitions Timeouts, retries and idempotency Tool result validation Human approval for high-impact actions Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools Workflow framework or explicit state machine Queue or task runner Structured logs
Hands-on exercise Model a multi-step assistant as a state machine and test repeated or failed steps. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project A workflow runner with bounded retries and approval checkpoints. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria The workflow is inspectable and cannot silently repeat sensitive actions. Evidence must be reproducible and tied to Generative AI.
Next: Security, privacy & cost engineering
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn PII minimisation and retention Secrets and tenant isolation Rate limits, caching and fallback models Prompt injection and output encoding Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools Secret manager Usage telemetry Policy tests
Hands-on exercise Threat-model a tenant-aware workflow and test cross-user isolation. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project A secure configuration and cost budget for the service. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria No secret or private user data appears in client logs or provider errors. Evidence must be reproducible and tied to Generative AI.
Next: Deployment & observability
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Model and prompt version pinning Trace spans and redacted logs Offline/online evaluation Rollback and graceful degradation Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools Monitoring platform CI pipeline Feature flags
Hands-on exercise Run a staged release with synthetic test cases and a rollback drill. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project A versioned service with dashboards and operational runbook. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria A failed model/dependency can be rolled back without losing user state. Evidence must be reproducible and tied to Generative AI.
Next: Capstone: grounded LLM application
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn User problem and data permissions Grounding and evaluation Security and operating cost Demonstration and architecture rationale Generative AI specialization: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Tools LLM API or local model Retrieval index if needed Evaluation harness
Hands-on exercise Present a live test set including failure and abuse cases and explain mitigations. Context for this route: Track-specific focus: Generative AI. Use the tools and standards named in this path; avoid treating this route as a generic overview of Artificial Intelligence.
Suggested project A deployable LLM application with tests, references, privacy notes and cost estimates. The outcome must directly demonstrate Generative AI, not merely list tools.
Learning resources Completion criteria The system is grounded, auditable and has explicit failure boundaries. Evidence must be reproducible and tied to Generative AI.
End of roadmap: Complete the final project and build your portfolio.
✓ Mark this node as complete