Content gaps: guide pillars and certifications¶

Last reviewed: 2026-11-07

A gap review of the guide pillars (learn/, work/, build/, tools/, safety/, careers/) and the certification guides on ai.noblerworks.com. Paper and explainer gaps stay in GAPS.md, which remains the queue for those. The review listed the tree with git ls-files and searched it with git grep. A topic is called missing only when the search found nothing (or one or two passing mentions). The open items are tracked in TODO.md under "Content gaps (2026-11-07 review)".

Where the site stands¶

The LLM half of the site is strong. The gaps are the classic-ML foundations a "zero" reader needs and the careers pillar.

Plan, in order¶

Wave 1 - finish what is written (already in TODO)¶

Run the untested project code (chatbot, extraction, voice agent, red-team, coding-agent loop, agent eval), the AI-103 notes and the GCP GenAI samples against the live APIs. A guide that fails when someone runs it costs more than a missing guide.

Wave 2 - foundations for the zero reader¶

• learn/day-one/: maths for ML without fear (vectors, matrices, probability, gradients: linear algebra has 2 passing mentions).
• learn/concepts/: decision trees and gradient boosting (3 mentions), CNNs and computer vision basics (3), reinforcement learning basics (mentioned often, never explained for beginners), time-series forecasting, feature engineering and data prep.
• build/projects/: a tabular ML end to end project in scikit-learn (train, evaluate, deploy), so the classic-ML pages have a hands-on partner.

Wave 3 - careers (the thinnest pillar)¶

• Roles: ML engineer roadmap, data scientist in the LLM era, AI product manager, AI researcher (the learn path exists; the career page does not).
• Building an AI portfolio (2 passing mentions): what to build, how to show it.
• Interviews: AI product manager, and prompt / applied-AI take-home exercises.

Wave 4 - work, tools and safety breadth¶

• work/playbooks/: HR, legal, finance, product and design.
• work/use-cases/: healthcare and education (2 mentions each, both regulated, so cite regulators).
• tools/comparisons/: image and video generation, speech (ASR/TTS) providers, embedding models, no-code AI automation (n8n, Zapier, Make: no hits).
• build/patterns/: multi-agent orchestration and guardrails as an architecture pattern.
• safety/: privacy and data protection with AI, AI incident response (one mention), deepfakes and synthetic media (one mention), bias and fairness testing in practice, US state AI laws (Colorado: one mention).

Wave 5 - certifications (already in TODO)¶

• NVIDIA NCP-AIN, NCP-OUSD, NCA-GENM rewrites to the blueprints (2026-11-15).
• Databricks ML Associate rebuild on the 4-section outline.
• Flashcard decks: generator ported 2026-11-07 (scripts/build_flashcards.py); CCDV-F and Databricks GenAI now have decks. Still no deck: NVIDIA ADS and GenAI LLMs Professional (waiting on the rewrite), Databricks ML Associate/Professional (add Key terms lines, then verify against the docs), retired AI-102 and DP-100, the prompt-engineering study track. Plus PMLE monitoring / Model Garden / Model Armor cards.
• New guides once codes are verified: AWS AI Business Strategist, NVIDIA NCA-ADS, NCP-ARI, MLA-C02 at GA. The AI governance certs (AIGP, AAISM) are planned for the security site with a link here.

How to do each item¶

Guide pages use templates/guide.md with the three header lines; run build_manifest.py, check_links.py and check_counts.py before committing. Prices and model names are cited to the vendor page with a date.