Get certified in
Applied AI & Machine LearningAssociate Readiness
Prove your ability to work with data, build and evaluate machine-learning workflows, and make sound entry-level ML decisions - across four parts, against a published competency blueprint.
Built for AI / ML Associate, Junior Data Scientist, Machine Learning Analyst and similar entry-level roles.
Join 100,000+ One Roadmap-certified candidates
The assessment
20 min
timed assessment
20
technical + scenario + case questions
4
assessment sections
75%
passing standard
You will be assessed on
Eight competencies, scored separately
Not one number called “AI/ML”. Your result reports a level for each of these, so you can pass overall and still see exactly where the gap is.
Machine Learning Fundamentals
Supervised and unsupervised learning, classification, regression, clustering and feature engineering.
Python for AI/ML
NumPy and pandas operations used in real data and modelling workflows.
Data Preparation
Missing values, duplicates, encoding, scaling and safe train/test preparation.
Model Building
Model fitting, regularization, ensemble methods and hyperparameter decisions.
Model Evaluation
Confusion matrices, precision, recall, F1, ROC/AUC and cross-validation.
Generalization & Reliability
Overfitting, underfitting, bias-variance and data leakage.
Model Selection & ML Judgment
Choosing an approach and a metric that fit the problem and its constraints.
Applied ML Problem Solving
Translating a real problem into a defensible ML workflow and spotting how it fails.
Assessment standard
Every attempt draws a fresh paper - but always the same number of questions from each competency area, so nobody gets an easier test.
AI/ML Core Knowledge
Learning types, algorithms, regularization, ensembles and the framing decisions that come before any of them.
- ML Fundamentals3
- Model Selection / Tuning2
- Applied ML Judgment2
Python & Data Preparation
pandas and NumPy code you have to read correctly, and the preparation decisions that decide whether a model is even trainable.
- Python / Pandas / NumPy3
- Data Preparation2
Model Evaluation & ML Judgment
Whether you can tell a good model from a good-looking number - and diagnose the difference.
- Model Evaluation3
- ML Failure Diagnosis2
You must score at least 60% here, whatever your overall total.
Applied ML Response
Three short written situations: what you would investigate, and what you would say.
- Applied ML Response3
Assessment version
AIML-ASSOCIATE-v2
Question bank
68 randomized items
Passing standard
75% weighted
Result bands
Distinction / Strong Performance / Certified
Result bands describe performance on this assessment. They are not a measure of employability or hiring outcomes.
Learning outcomes
These are what the public verification page reports - and each one is claimed only where your own attempt evidenced it.
Machine Learning Foundations
Demonstrated understanding of supervised and unsupervised learning, classification, regression, clustering and common ML approaches.
Python for Data & ML
Demonstrated ability to reason through common pandas and NumPy operations used in data preparation and machine-learning workflows.
Data Preparation
Demonstrated understanding of missing-value handling, duplicates, feature encoding, scaling and safe train/test preparation.
Model Building
Demonstrated understanding of model training, regularization, ensemble methods and common model configuration decisions.
Model Evaluation
Demonstrated ability to interpret precision, recall, F1, ROC/AUC, confusion matrices and cross-validation results.
Model Reliability
Demonstrated understanding of overfitting, underfitting, data leakage, bias-variance trade-offs and validation strategy.
Applied ML Judgment
Demonstrated ability to select appropriate modeling approaches and evaluation metrics based on the problem and business objective.
Applied ML Problem Solving
Demonstrated ability to translate real-world problems into defensible machine-learning workflows and identify common failure modes.
How the written responses are graded
A model reading your answer and returning “8.7 out of 10” is not a grade - it is a guess with a decimal point. So we split the job in two: the AI only observes, and the code scores.
The evaluator reports which technical behaviours it can evidence in your own words - identified overfitting, checked for leakage, tied the metric to the objective. It has no score field to write into. Points come from a fixed rubric in code, so correct ML vocabulary with no reasoning behind it earns nothing.
Your written response
50-100 words, in your own words, in the time you have.
AI evaluator
Reports only which expected behaviours are evidenced, and which red flags appear.
Deterministic rubric
Code turns those observations into points, using the bank's own signal list.
Weighted result + competency profile
Part A 35%, Part B 25%, Part C 25%, Part D 15% decides the 75% pass mark; the same answers roll up into a level for each competency.
Weekly Ranking
Frequently asked questions
Free, 20 minutes, and a verifiable credential at the end.
This credential assesses machine-learning and data-science foundations. It does not verify modern LLM or generative-AI engineering - RAG, embeddings, prompt and model evaluation, agents, structured generation or production AI systems - and is deliberately not positioned as an AI Engineer credential.