Find the words that decide the architecture, then decide it yourself.
Built to teach SAA-C03 candidates and cloud engineers to turn scenario constraints into architecture decisions, as an accepted browser-only SAA-C03 prototype.
The idea
CertGraph is for two groups: people preparing for SAA-C03, and cloud engineers who want sharper AWS architecture judgment. Both hit the same wall. A long scenario question hides a few phrases that decide the answer, and much practice rewards remembering which option was right rather than seeing why the others fall short. Often the gap is not knowing the services but misreading a constraint.
When it works, CertGraph feels like studying beside an architect who thinks out loud. You watch a scenario become a short set of constraints, see every option checked against them, then do the same work yourself as the help fades from worked example to coaching to solo. The bet is that generic AI has made explanations and question banks cheap, while trustworthy reasoning practice, with cited sources and visible tradeoffs, stays hard to build.
Further out, the direction is the reasoning layer between technical learning and technical operations, moving from deeper AWS practice toward team diagnostics, with each step earned by evidence before the next one starts. Today CertGraph is narrower: an accepted browser-only SAA-C03 prototype with 88 synthetic practice items, free and with no account. Independent AWS review and a learner study are still ahead, so it makes no claim yet about results.
How it works
Model, Coach, Solo
Pattern
Mock
Architecture Lab
Each scenario's phrases, constraints, option impacts, and decision are stored as explicit, versioned, source-cited content, so a correction creates a new version rather than rewriting past attempts. Learning runs and saves in the browser, and the prepared guidance is not a live AI tutor. The only server path is opt-in, pseudonymous analytics and feedback, which never include answers, scores, or reasoning. No protected assessment items are stored or republished.
Where it stands today
An accepted, browser-only SAA-C03 learning prototype, free to use with no account. It has 10 phrase-level scenarios worked through Model, Coach, and Solo, plus 48 Pattern drills, three coaching-free 10-question Mock forms, and 10 Architecture Lab challenges. All of this content is synthetic, cited to official AWS documentation, and was revised and re-audited with agent assistance in September 2026. Learning state stays in the browser, and the only server path is opt-in, pseudonymous product analytics and a feedback form; independent AWS review and learner validation are still open.
Working now
Model, Coach, and Solo deconstruction across 10 scenarios: find the deciding phrases, turn them into constraints, and test every option, with help that fades
Option reasoning made visible: each option marked against each constraint, with authored rationales and official AWS source links
Pattern Trainer: 48 drills in 6-, 12-, or 24-question sessions, with an explained recommended mode and an optional timer
Three coaching-free, 10-question Mock forms with per-option rationales and links back to targeted practice
Architecture Lab: build a topology for each scenario and check it against that scenario's rules
Browser-only progress with export, plus opt-in pseudonymous analytics and a feedback form
Not there yet
Independent AWS subject-matter review has not been completed.
No learner study or validated evaluation set exists yet, so whether the method transfers to unseen questions is unknown.
How it got here
The graph line starts in myDNAHealth, whose private results graph let a person explore their own matches
AWS Glossary Graph puts AWS vocabulary into a public, explorable graph
Broad learning-platform experiments, including a small force graph
CertGraph started, first as a draft inside that learning-platform work
Current product imported and re-anchored on phrase-level deconstruction, with Architecture Lab kept separate
Learning loop closed and published: Model, Coach, and Solo, then Pattern, Mock, and targeted practice
Product acceptance completed
A prelaunch review paused the broad announcement until two correction passes shipped
All 88 practice items revised and re-audited
What comes next
Next step
Complete independent AWS technical review and a bounded discovery round with SAA-C03 learners before making any efficacy claim.
Later ideas, not current features
Complete independent AWS technical review of the 88-item bank
Run a small, bounded discovery round with SAA-C03 learners and read the opt-in product signals
Prepare, but not yet start, a learner study that tests whether the method transfers to unseen questions
What I checked, and what stays bounded
Dated evidence
Product acceptance completed for the browser-only SAA-C03 prototypeacceptance
Synthetic-content, source-backing, and independent-review gates recordedgovernance
Public certification scope re-verified against the canonical SAA-C03 repositorycorrection
Opt-in product signals and feedback went live, with no answers or reasoning collecteddeployment
All 88 practice items revised and re-audited against AWS documentation (agent-assisted, not independent review)correction
Latest release verified in production: first-run CI and a production containment check passeddeployment
Boundaries it keeps
No actual, recalled, paid, or protected AWS assessment items are republished
Practice content is synthetic and agent-assisted, and cited to official AWS documentation
Product acceptance is not evidence of improved learning outcomes
No learner study has been run, and no efficacy, retention, or pass-rate claim is made
Opt-in product signals count browser installations, not people, and no current learner number is claimed
CertGraph does not access AWS accounts, credentials, or live cloud resources, and it operates no one's infrastructure
Prepared, deterministic guidance; no live AI tutor
Not affiliated with AWS; mock results are not official exam scores
My role
Set the learning thesis, the SAA-C03 scope, the Model, Coach, and Solo method, the synthetic-content and source standards, and the acceptance criteria; directed, reviewed, and tested the agent-assisted implementation and the 88-item bank revision, and decided what shipped.