How to score yourself
Use a four-point evidence scale. Do not average scores blindly: strong AI theory with no production software, security or stakeholder delivery still represents critical readiness gaps.
| Score | Meaning | Evidence standard |
|---|---|---|
| 0 | Not yet exposed | Cannot explain or demonstrate it |
| 1 | Understand | Can explain concepts and identify examples |
| 2 | Apply with support | Has completed a guided lab or bounded task |
| 3 | Demonstrate independently | Has produced inspectable evidence and can defend trade-offs |
The ten-domain checklist
1. Discovery & qualification
- Map stakeholders, users, decisions and exceptions.
- Separate a root problem from a feature request.
- Establish a KPI baseline and measurement source.
- Recommend “do not use AI” when appropriate.
2. Business-to-technical translation
- Write a testable problem and target outcome.
- Define approvals, fallbacks and exception routes.
- Create acceptance criteria and an evaluation plan before build.
3. Production software engineering
- Build and review Python services and a modern interface or integration.
- Design APIs, SQL data and failure handling.
- Use tests, authentication, authorization and secure defaults.
4. Data & knowledge engineering
- Inventory sources, owners, classifications and quality.
- Design governed ingestion, retrieval, lineage and freshness.
- Evaluate retrieval independently from answer quality.
5. LLM, RAG & agentic AI
- Select models using quality, latency, cost and privacy.
- Validate structured outputs and evaluate RAG.
- Bound tools, state, permissions, approvals and escalation.
6. Enterprise integration
- Connect APIs, events, SaaS and legacy systems safely.
- Handle identity, delegated authorization, retries and idempotency.
- Design for degraded dependencies and auditability.
7. Evaluation & quality
- Create representative, adversarial and regression cases.
- Define metrics, segments, thresholds and release gates.
- Calibrate human and model-assisted evaluation.
8. Security & governance
- Threat-model prompt injection, leakage and excessive agency.
- Apply least privilege, human oversight and audit controls.
- Document residual risk and named accountability.
9. Cloud & AI operations
- Separate environments and gate release with CI/CD.
- Instrument application, model and tool traces.
- Rehearse rollback, fallback and incident response.
10. Adoption, value & communication
- Run UAT around real roles and tasks.
- Distinguish availability, adoption and value.
- Present trade-offs to engineers, operators and executives.
Minimum viable FDE portfolio
One complete case study is stronger than six disconnected demos. Include the problem and baseline, current and target workflow, architecture, code evidence, governed data, evaluation dataset, threat model, CI/CD, rollback, observability, UAT, adoption, KPI evidence, individual contribution, failures and a reproducible verification guide.
Evidence quality test
Authentic: Did you materially contribute? · Inspectable: Can a reviewer verify more than a screenshot? · Contextual: Does it explain the decision? · Measured: Is there a baseline or threshold? · Safe: Is confidential information protected? · Defensible: Can you explain alternatives and failures?
Redaction and confidentiality
Never publish credentials, production data, personal information, proprietary code, customer names without permission, internal URLs or sensitive security architecture. Use synthetic data, redacted artifacts, private review or a reproducible simulation. A portfolio is evidence of judgment; exposing confidential information demonstrates poor judgment.
Interpret your result
| Pattern | Recommended next step |
|---|---|
| Mostly 0–1 in engineering foundations | Build programming, SQL, API, Git and cloud foundations |
| Strong anchor; weak AI/evaluation | Add focused applied-AI and evaluation work |
| Strong AI; weak software/operations | Deploy a tested service with CI/CD and observability |
| Strong technology; weak discovery/adoption | Lead a stakeholder-facing deployment simulation or pilot |
| Mostly 2–3 with one clear anchor | Consider advanced FDE preparation |
This is developmental guidance, not an employment or admissions guarantee.
Turn your scores into a bridge plan.
Bring your strongest evidence and biggest gaps to a short advisor conversation.
Talk to an advisor →