Curriculum / AI Engineer
AI Engineer — AIP Builder
Ship a governed AIP decision-support capability for disruption response: an AIP Logic recommendation function (Use LLM + Ontology tools), a code-authored TypeScript LLM tool with structured output, model governance (BYOM register + scope), an AIP Evals suite, an AIP Chatbot Studio 'Disruption Copilot' with safe tools, and production governance (permissions, Markings, staged edits, audit) — culminating in an integrated, evaluated, governed copilot.
AI engineers and advanced analysts (coding-capable — AIP Logic + TypeScript Functions).
The 8-unit arc
- Unit 1: Frame the governed decision: a single Use LLM Logic blockintermediate
Build a saved AIP Logic function (recommendDisruptionResponse) that takes a DisruptionEvent and returns a structured recommendation via one Use LLM block.
- Unit 2: Make the recommendation reliable: multi-block Logic with Ontology toolsintermediate
Expand the function with a Query objects tool over Shipment/Supplier, an exposure calculation, and a Conditional that branches recommendation severity.
- Unit 3: Author an LLM tool in code: TypeScript Functions + structured outputadvanced
Publish a TypeScript Function (classifyDisruptionSeverity) that calls a model's createChatCompletion and returns a structured typed result, usable as a Call function tool.
- Unit 4: Govern the model: register a model (BYOM) and scope itadvanced
Register a model (BYOM), declare its capabilities, select it for the Logic/TS function, and review rate limits — recording the defense caveat that BYOM models don't support Markings.
- Unit 5: Prove it before release: AIP Evals suiteadvanced
Build an AIP Evals suite over recommendDisruptionResponse with multiple disruption test cases and an evaluator, run with a metrics report.
- Unit 6: Expose it conversationally: AIP Chatbot Studio with safe toolsadvanced
Publish a 'Disruption Copilot' chatbot with an Object query tool over the disruption Ontology and a Function tool calling the published recommendation function, using native tool calling.
- Unit 7: Wrap it in governance: permissions, Markings, staged edits, auditadvanced
Deliver the recommendation's Ontology edit via a governed Action (submission criteria / staged-for-review, not auto-applied), minimize tool scopes, and map the workflow to responsible-AI themes.
- Unit 8: Capstone: integrated, evaluated, governed decision-support releasecapstone
The end-to-end Disruption Copilot: published function + eval suite + chatbot + governed Action, demonstrated on a seeded disruption with an auditable recommendation-to-edit trace.
Frame the governed decision: a single Use LLM Logic block
Build a saved AIP Logic function (recommendDisruptionResponse) that takes a DisruptionEvent and returns a structured recommendation via one Use LLM block.
Make the recommendation reliable: multi-block Logic with Ontology tools
Expand the function with a Query objects tool over Shipment/Supplier, an exposure calculation, and a Conditional that branches recommendation severity.
Author an LLM tool in code: TypeScript Functions + structured output
Publish a TypeScript Function (classifyDisruptionSeverity) that calls a model's createChatCompletion and returns a structured typed result, usable as a Call function tool.
Govern the model: register a model (BYOM) and scope it
Register a model (BYOM), declare its capabilities, select it for the Logic/TS function, and review rate limits — recording the defense caveat that BYOM models don't support Markings.
Prove it before release: AIP Evals suite
Build an AIP Evals suite over recommendDisruptionResponse with multiple disruption test cases and an evaluator, run with a metrics report.
Expose it conversationally: AIP Chatbot Studio with safe tools
Publish a 'Disruption Copilot' chatbot with an Object query tool over the disruption Ontology and a Function tool calling the published recommendation function, using native tool calling.
Wrap it in governance: permissions, Markings, staged edits, audit
Deliver the recommendation's Ontology edit via a governed Action (submission criteria / staged-for-review, not auto-applied), minimize tool scopes, and map the workflow to responsible-AI themes.
Capstone: integrated, evaluated, governed decision-support release
The end-to-end Disruption Copilot: published function + eval suite + chatbot + governed Action, demonstrated on a seeded disruption with an auditable recommendation-to-edit trace.