Skill matrix.
Use this as a prioritization board: what each skill area requires at production level, and how to prepare it.
| Skill area | What "production-ready" requires | Priority | How to prepare |
|---|---|---|---|
| Enterprise Gen AI architecture | Define a foundation architecture and technical roadmap for a Gen AI platform. | P0 | Prepare a 2-minute reference architecture and a 6-month roadmap narrative. |
| RAG architecture | Design and reason about production retrieval-augmented generation systems. | P0 | Study ingestion, chunking, embeddings, hybrid retrieval, reranking, citations, and evaluation. |
| LLM fine-tuning | Know when and how to fine-tune versus retrieve versus prompt. | P1 | Prepare a RAG-vs-fine-tuning-vs-prompting decision framework with concrete tradeoffs. |
| LLMOps | Operate prompts, models, and evidence as a running production system. | P0 | Learn prompt/model registries, eval harnesses, drift detection, rollback, and cost/latency monitoring. |
| Evaluation | Define and measure quality before and after shipping. | P0 | Practice designing a golden dataset and a scoring rubric for a given use case on the spot. |
| Guardrails and safety | Design input/output controls that hold under adversarial and edge-case input. | P0 | Study prompt-injection defense, output scanning, and human-in-the-loop gates for irreversible actions. |
| Agentic systems | Bound autonomy, tool access, and state in multi-step agent workflows. | P1 | Study control-loop termination conditions, tool-permission scoping, and failure-mode design. |
| Security, privacy, governance | Apply access control, PII handling, and regulatory-risk thinking to AI systems. | P1 | Study risk tiering, audit logging, and the difference between accepted risk and a hard blocker. |
| Cost architecture | Reason about unit economics and budget planning for a Gen AI platform. | P2 | Practice building a Bill of Materials and unit-economics model for a sample system. |