Free Practice Questions for Palantir AI Engineer Associate Certification

Subdomain 1.1: Configure typed inputs and outputs, system and task prompts, variable injection, tools, blocks, conditionals, and loops.

Subdomain 1.2: Choose single completion or tool-using reasoning appropriately.

Subdomain 1.3: Apply Ontology edits safely and understand when code is more suitable than AIP Logic.

Subdomain 1.4: Debug traces, tune parameters, publish versions, and preserve behaviour for existing consumers.

Subdomain 2.1: Distinguish Standard Agents and AIP Assist based on deployment context and capabilities.

Subdomain 2.2: Configure tools, retrieval context, conversation behaviour, guardrails, and application state.

Subdomain 2.3: Use Workshop embedding, standalone access, and AIP Studio appropriately, including multi-agent coordination where required.

Subdomain 2.4: Test representative, edge, and adversarial interactions before production release.

Subdomain 3.1: Design chunking, metadata, embedding, vector-property, semantic-search, and context-injection strategies.

Subdomain 3.2: Use the same embedding model for indexing and query-time retrieval.

Subdomain 3.3: Balance precision, recall, freshness, token limits, and latency.

Subdomain 3.4: Diagnose whether poor answers originate in chunking, embeddings, ranking, thresholds, context selection, or generation.

Subdomain 4.1: Configure batch LLM processing, entity extraction, classification, summarisation, visual document processing, structured output, and validation.

Subdomain 4.2: Preserve prompts and raw responses where auditability and debugging require them.

Subdomain 4.3: Decide whether computation belongs south of the Ontology for reusable batch enrichment or north for interactive, context-dependent reasoning.

Subdomain 4.4: Codify workflows in Code Repositories when scheduled batch execution, programmatic control, rate limiting, or broader libraries are required.

Subdomain 5.1: Create representative evaluation suites, test cases, target functions, evaluators, metrics, and experiments.

Subdomain 5.2: Test normal, edge, adversarial, unauthorised, and high-consequence cases.

Subdomain 5.3: Distinguish signal from non-deterministic variance and detect regressions across prompts, models, and versions.

Subdomain 5.4: Design feedback loops that capture user corrections or downstream outcomes and turn them into new evaluation cases or improvement priorities.

Subdomain 6.1: Configure Automate triggers, effects, execution modes, function-backed Actions, permissions, and Workshop integration.

Subdomain 6.2: Chain Functions, Actions, and AI outputs into multi-step workflows.

Subdomain 6.3: Test manually with representative objects before enabling live triggers.

Subdomain 6.4: Modify live workflows safely, accounting for retries, duplicate execution, partial failure, and in-flight work.

Subdomain 7.1: Select a model from the supplied reference characteristics using capability, latency, cost, geography, and task requirements.

Subdomain 7.2: Understand multi-provider resilience, BYOM, and when a custom endpoint is appropriate.

Subdomain 7.3: Use AIP Observability, execution history, session traces, distributed traces, logs, token telemetry, and cost signals.

Subdomain 7.4: Optimise prompts, context, model choice, caching, and execution structure without sacrificing quality or safety.

Subdomain 8.1: Apply enrolment and capability permissions, execution security modes, georestrictions, and capacity controls.

Subdomain 8.2: Understand audit and lineage evidence, data-protection architecture, and sensitive data handling.

Subdomain 8.3: Add human review for high-consequence actions and match validation strictness to consequence.

Subdomain 8.4: Recognise prompt injection, jailbreak, data exfiltration, and unsafe tool-use risks and select defensive controls.