This plan is built around the fundamentals loop: classify the task -> choose the metric -> check failure modes -> route governance and monitoring.
How to use this plan well
Each study block should do four things:
- classify the issue as data, evaluation, GenAI, governance, or lifecycle
- choose the metric, control, or concept that belongs there
- do a short scenario or mixed set
- write down whether the miss came from data quality, wrong metric, weak grounding, or ignored governance
flowchart LR
Classify["Classify data / evaluation / GenAI / governance / lifecycle"] --> Choose["Choose metric or control"]
Choose --> Drill["Do short scenario set"]
Drill --> Review["Review why misses happened"]
Review --> Classify
How long should you study?
Typical candidates need 25 to 55 focused hours.
| Your time |
Recommended timeline |
Good fit |
| 10 to 12 hrs/week |
30 days |
intensive path with some prior AI exposure |
| 5 to 7 hrs/week |
60 days |
balanced path for most candidates |
| 2 to 4 hrs/week |
90 days |
part-time path with slower reinforcement |
30-day intensive plan
| Week |
Focus |
Output |
| 1 |
AI and ML basics, problem framing, data quality, and task types |
fundamentals notes and short drills |
| 2 |
metrics, leakage, imbalance, overfitting, and evaluation design |
metric tie-break sheet |
| 3 |
GenAI basics, grounding, retrieval, and hallucination control |
weak-lane notes and mixed sets |
| 4 |
responsible AI, privacy, fairness, monitoring, and final compression |
mixed review and readiness check |
60-day balanced plan
| Phase |
Weeks |
Focus |
| 1 |
1 to 2 |
lifecycle basics, task framing, and data-quality thinking |
| 2 |
3 to 4 |
metrics, leakage, overfitting, and evaluation logic |
| 3 |
5 to 6 |
GenAI basics, grounding, retrieval, and answer-quality judgment |
| 4 |
7 |
responsible AI, privacy, fairness, and monitoring |
| 5 |
8 |
weak-lane repair and mixed review |
90-day part-time plan
| Month |
Focus |
Goal |
| 1 |
vocabulary, task framing, and lifecycle basics |
stop losing points to term confusion |
| 2 |
data and evaluation logic |
get stronger at metric and failure-mode judgment |
| 3 |
GenAI, governance, and exam-style tie-breaks |
finish with mixed-set confidence |
If misses cluster here, do this next
| Miss pattern |
Weak lane |
Fix next |
| you choose the wrong score or trust accuracy too fast |
evaluation |
review precision, recall, F1, AUC, and task-fit logic |
| you trust great offline results too easily |
data and evaluation |
review leakage, split design, and held-out validation |
| you blame the model when the context is weak |
GenAI |
review grounding, retrieval quality, and supported answers |
| you skip fairness, privacy, or safety |
governance |
review responsible AI controls and operational ownership |
What strong prep usually does
- classifies the problem before naming a technique or product
- keeps confusion tables for leakage vs overfitting, accuracy vs F1, and grounding vs fine-tuning
- writes down why the winning answer is safer or more aligned with the task instead of memorizing AI buzzwords
- uses Oracle docs to settle disagreements, then comes back here for compression
Final 72 hours
| Keep doing |
Stop doing |
| rereading the cheat sheet and glossary |
opening unrelated new AI tools or framework tutorials |
| reviewing weak-lane notes and confusion pairs |
treating every question like a product-feature question |
| checking Oracle docs for disputed boundaries |
building a large late-stage AI lab |
| practicing data -> metric -> model -> governance order |
trusting unsupported community summaries over Oracle docs |
Route yourself well
- last-mile metrics, GenAI, and safety traps: Cheat Sheet
- last-week questions: FAQ
- high-confusion AI fundamentals terms: Glossary
- official Oracle and OCI sources: Resources