Maple 2: evaluate a canary without hiding version failures
Advanced deployment lab · Use per-version denominators, predeclared gates and explicit holds for insufficient evidence.
Existing companion · depth review pendingOpen collection →Read deeply.
Put your understanding to work.
Explore a topic as a collection of learning resources. Choose a chapter, work through a question, inspect the data or return to an explanation.
Explore the SQL reference collection →201 topic collections. Page 10 of 17. Existing companions are not automatically complete study units.
Maple 3: monitor service health and model quality separately
Advanced MLOps review · Treat missing labels as unknown quality, distinguish drift from degradation and define a monitoring contract.
Existing companion · depth review pendingOpen collection →Maple 4: rehearse rollback and write an AI deployment review
Advanced portfolio and incident exercise · Connect release evidence, owner roles and recovery limits without overstating a local simulation.
Existing companion · depth review pendingOpen collection →Iris 1: define the retrieval and answer-evaluation contract
Advanced AI evaluation · Separate relevance, evidence support, answerability and user decisions before choosing a metric.
Existing companion · depth review pendingOpen collection →Iris 2: evaluate ranked retrieval with honest denominators
Advanced retrieval lab · Calculate cutoff metrics, freeze development selection and inspect multi-evidence misses.
Existing companion · depth review pendingOpen collection →Iris 3: audit evidence support and abstention together
Advanced AI review lab · Inspect atomic claims, source applicability and withheld answers without treating labels as an automated judge.
Existing companion · depth review pendingOpen collection →Iris 4: defend an AI evaluation and release recommendation
Advanced portfolio and interview exercise · Turn retrieval failures, evidence reviews and test limitations into a clear decision memo.
Existing companion · depth review pendingOpen collection →Elm 1: define the causal question and inspect confounding
Advanced causal reasoning · Distinguish an observed association from an intervention effect before adjusting a table.
Existing companion · depth review pendingOpen collection →Elm 2: evaluate assignment without dropping non-compliers
Advanced experiment lab · Preserve the assignment estimand and enumerate a small exact randomization test under an explicit design.
Existing companion · depth review pendingOpen collection →Elm 3: inspect difference-in-differences assumptions
Advanced policy-review lab · Compare before-after, post-only and DiD contrasts, then challenge the missing counterfactual trend.
Existing companion · depth review pendingOpen collection →Elm 4: defend a causal conclusion with an assumption ledger
Advanced portfolio and interview exercise · Communicate an estimand, design, evidence and sensitivity without turning a contrast into a causal promise.
Existing companion · depth review pendingOpen collection →Cedar 1: define the forecast before choosing a model
Advanced forecasting · Separate the target, issue time, horizon and available information before comparing methods.
Existing companion · depth review pendingOpen collection →One useful next step
Start learning. Build understanding.
Explore the introductory learning, then create a free account to continue, save learning notes and track supported lessons. Registration and email verification cost nothing.