IFoA IFOA_CAA_M0 Exam Prep
Certified Actuarial Analyst Exam (Page 3 )

• Fine-tuning requires a properly formatted training dataset.
• For classification or instruction-style fine-tuning, examples are supplied in JSONL format.
• Each line represents one training example, typically containing the input prompt and the desired completion or label.
• The file must then be uploaded or made available to the Microsoft Foundry fine-tuning workflow.

• A: Prompt Flow is useful for prompt experimentation, but it is not required to start fine-tuning.
• B: You do not need to deploy the base model to an online endpoint before fine-tuning.
• C: Tracing supports observability and evaluation, but it does not prepare the fine-tuning job or dataset.

• Suspended an operator who correctly refused to use unsafe equipment.
• Rewarded another operator for using the equipment despite the broken safety guard.

• A — Management is not particularly concerned with managing risks.

• B — Employees will perceive that management does not care about health and safety risks.

• C — This factory’s safety record is likely to be poor.

1. Write a structured log entry to CloudWatch Logs containing:

1. Create a CloudWatch Logs metric filter that:

text 
ApiRequests 
Operation=GetOrders 
ResponseCode=500 
Version=2.4.0

• B is incorrect: CloudWatch Logs Insights queries analyze logs but do not directly publish metrics in this manner.
• C is incorrect: ALB access logs do not contain arbitrary Lambda response metadata as custom fields.
• D is incorrect: X-Ray traces and insights are not the appropriate mechanism for creating these custom, dimensioned application metrics.

• Lower order numbers appear first.
• For example, a variable with order 100 appears before one with order 200.
• The Sequence field is not the control used for catalog variable display order.
• The Default Value sets a value, not placement.
• The Choice field defines selectable options, not the variable’s position.

• Report on the discount approval time for quotes

• Dashboard of opportunity stage duration

• Lead conversion trends measure lead-generation effectiveness, not opportunity closing time.
• Campaign ROI evaluates marketing performance, not delays in the sales process.

• Economy: Obtaining inputs at the lowest appropriate cost.
• Efficiency: Making the best use of resources, such as staff, money, and equipment.
• Effectiveness: Achieving the intended objectives or outcomes.

• Azure DevOps Analytics views are accessed through the Power BI Azure DevOps (Boards only) connector.
• This connector authenticates directly to Azure DevOps Services and lets you select an Analytics view.
• Analytics views allow filtering and shaping data in the cloud before Power BI loads it.
• Azure DevOps Server (Boards only) is for an on-premises Azure DevOps Server instance, not Azure DevOps cloud.
• A generic OData Feed connection can access OData endpoints, but it is not the purpose-built connector for Azure DevOps Analytics views.
• “OData queries” describes the underlying technology or query approach, not the most appropriate Azure DevOps connector choice here.

• Custom domains: Supported by Basic and higher tiers.
• Dedicated compute: Basic and Standard use dedicated virtual machine instances; Free and Shared use shared infrastructure.
• Load balancing between instances: Standard provides scaling across multiple instances with load balancing.
• Storage: Standard provides more storage than Basic.
• Cost: Standard is the least expensive option that satisfies all the stated requirements.

• Free/Shared: Use shared compute and have limited features; they don’t meet the dedicated-compute requirement.
• Basic: Provides dedicated compute and custom domains, but does not provide the required Standard-level multi-instance capabilities in the exam’s feature comparison.

• A. Continue using the current system: It will eventually fail and offers no recovery strategy.
• B. Virtualize the entire system: Correct. A VM can preserve the old OS, application, settings, and dependencies while running on newer physical hardware.
• C. Containerize the application: Containers generally share the host OS kernel and are not suitable for preserving an entire incompatible legacy OS.
• D. Clone the disks to USB: A disk clone may provide backup or temporary recovery, but it does not solve hardware compatibility or provide a reliable long-term platform.

• File is shared externally ? Access level

• File is labeled “Internal only” ? Sensitivity label

• Which batches and markets are affected
• Whether the deficiency affects product quality, efficacy, or patient safety
• Whether the issue is an ongoing manufacturing problem
• Whether immediate containment or distribution suspension is needed

• A — Withdraw the product: may be necessary, but is premature before assessing the risk.
• B — “Dear Dr.” letter: may be appropriate later, depending on the risk.
• C — Notify global authorities: may become mandatory, but the scope and urgency depend on the assessment.

• Process Builder: Automates actions based on record changes; it does not create interactive user forms.
• Approval Process: Routes records for approval and captures approval decisions.
• Outbound Message: Sends information to an external system; it does not provide a user interface.

• Existing expected ROI: 15%
• New project ROI: 14%
• Cost of capital: 12%

• Project return = 14%
• Cost of capital = 12%
• It earns 2% more than the cost of capital, so it should increase shareholder value.

• CBWFQ (Class-Based Weighted Fair Queuing): Creates traffic classes and assigns each class a guaranteed bandwidth or queueing treatment.
• PQ (Priority Queuing): Uses multiple queues with strict priority. Higher-priority queues are serviced before lower-priority queues, making it useful for delay-sensitive traffic—but it can starve lower-priority traffic.

• FRTS: Traffic shaping for Frame Relay; it controls the transmission rate rather than selecting packets during congestion.
• CAR: Polices traffic by rate, typically dropping or remarking packets that exceed a limit.
• PBR: Policy-Based Routing; it selects forwarding paths and is not a QoS queueing mechanism.

• On-premises resources, such as an existing database
• Cloud resources, such as Azure services or virtual machines
• Connectivity that allows the application to use both environments

• Public cloud: resources are entirely provided through a public cloud provider.
• Private cloud: resources are dedicated to one organization, typically on-premises or privately hosted.

• User-generated content: precisa ser coletado, filtrado, anonimizado e rotulado.
• Moderation logs: podem conter dados inconsistentes, informações pessoais e decisões que precisam ser revisadas.
• Content moderation guidelines: são regras ou orientações, não dados diretamente utilizáveis para medir o comportamento do modelo.

• COE: Identifies the Center of Excellence, such as Payroll or Benefits.
• Topic Detail: Classifies the HR service within its COE; it does not identify the catalog item.
• Template: Supplies default values or case configuration.
• Record Producer: Represents the catalog item that employees use to submit the request and create the case.

• The audience and stakeholders
• The complexity and type of requirements
• The communication method and level of detail needed
• Whether to use diagrams, models, documents, demonstrations, or slides

• B: Authority to host the presentation is not the key preparation activity.
• C: Stakeholder sign-off may occur later; presenting requirements does not always require formal approval beforehand.
• D: A scribe may be useful in a meeting, but hiring one is not a core preparation requirement for presenting.

• Record changes — a flow can start when a record is updated or otherwise changes.
• Schedules — a flow can run at a specified date/time or on a schedule.
• Record inserts — a flow can start when a new record is created.

• Subflows are reusable flow components called by another flow; they are not normally trigger types.
• Record views do not generally initiate a Flow Designer flow.

• The requirement is an assessment of the Azure environment’s design, which requires a support plan offering proactive architectural guidance.
• Premier support would provide this capability, but it is a high-cost option.
• Since the requirement says to minimize expenses, Premier is not the best recommendation.
• A lower-cost plan with the required advisory/design-review capability—historically Professional Direct in Azure’s support-plan model—would be more appropriate.

Why this is correct

• Each user authenticates with their own Google account.
• Google manages the user’s SSH key association.
• IAM controls who can access instances.
• Access can be audited by user identity, so security can determine who accessed a VM.
• Granting the role to a Google group avoids managing permissions individually.

Why the others don’t apply

• A: A shared private key is difficult to distribute securely and cannot reliably identify which employee used it.
• B: Manually collecting and deploying keys on every instance is operationally inefficient and complicates auditing.
• D: A project-wide shared key has the same identity and key-management problems as option A. It also does not provide per-user accountability.

• Data only needs to be current at 7 AM each day, so scheduled refresh is sufficient.
• Import mode provides the fastest report interaction because data is stored in Power BI’s in-memory engine.
• It supports row-level security (RLS), including the USERNAME() filter.
• You can limit the imported data to the current year and previous three years to reduce model size.
• DirectQuery would generally provide slower visual response times and is unnecessary here.
• Dual is mainly useful in composite models where some tables use DirectQuery and others use Import. That is not required by this scenario.
• Live connection is not appropriate because you are creating the semantic model.

1. Which host originated the infection
2. Whether each other host is clean or infected
3. Which firewall or host logs provide evidence of the infection’s movement

• Find the earliest infection-related event. The host associated with that event is the likely origin.
• Follow the timeline: outbound connections from that host may indicate propagation or command-and-control activity.
• A host receiving suspicious traffic is not automatically infected; look for evidence that it executed or initiated malicious activity.
• A firewall log can show traffic paths, but endpoint logs are usually needed to confirm infection.
• Mark a host clean only when the logs show no indicators of compromise or malicious activity.

• A — Improved management of storage space: Correct.

• B — Improved supplier relationships: Correct.

• C — Improved pricing decisions: Correct.

• D — Non-product-specific costs must be apportioned: Incorrect.

• E — Profitability of each customer group: Incorrect.

• Sales processes: Define the available Opportunity stages for each team.
• Record types: Connect each team to its appropriate sales process and page layout.
• Page layouts: Control which fields and related sections each team sees.

• Solar market team 1
• Solar market team 2
• Installation Services team

• One sales process: Cannot provide different Opportunity stages for each team.
• One page layout: Cannot provide different field arrangements or fields by team.
• Three page layouts alone: Page layouts control fields, but not the available stages.

Step 1: Calculate the rights issue price

Step 2: Calculate the number of new shares

Step 3: Calculate the value after the issue and project

• Rights issue proceeds: $21.6m
• Project NPV: $5m

Step 4: Calculate the theoretical ex-rights price

• The connected switch
• The switch’s port/interface
• Device and port descriptions
• Sometimes VLAN or management information

• IKE: Establishes security associations for IPsec VPNs.
• VLAN: Defines logical network segmentation; it does not identify a physical switch port.
• netstat: Displays local network connections and statistics, not the upstream switch and port.

• A. Remove rows where Sales[Status] is Canceled

• E. Remove Sales[Canceled Date]

• B: Removing Sales_Date would remove information users need—the date portion.
• C: Changing Delivery_Time to Integer could alter its meaning or precision and is not necessarily safe.
• D: Splitting the date/time column creates additional columns rather than simply removing unnecessary data.

• Internal hostnames and subdomains
• Network structure and naming conventions
• Servers and services
• Relationships between systems

• Determining ISP-assigned address space: Usually involves external registration and DNS information, not internal host discovery.
• Bypassing DNS sinkholing: Would require evidence that malicious DNS requests avoided redirection.
• Initial access: DNS queries alone do not show that the attacker gained access.
• Exfiltrating data: DNS tunneling can do this, but the snippet would typically show encoded or unusually long DNS queries.

• Response and resolution times
• Availability or uptime
• Performance targets
• Support obligations
• Penalties for failing to meet commitments

• SOW (Statement of Work): Describes the work, deliverables, scope, and tasks.
• MOA (Memorandum of Agreement): Documents a formal agreement about roles or responsibilities.
• MOU (Memorandum of Understanding): Records a general mutual understanding, often without detailed service metrics.

• Streams — track change data for tables or views.
• External tables — reference data stored in an external cloud storage location.

• Role — account-level security object.
• Warehouse — account-level compute resource.
• User — account-level security object.
• Share — account-level object used to share data.

• A user connects with their Google account.
• Google associates the user’s SSH key with that identity.
• IAM roles determine whether they can log in.
• Access can be tied to a specific employee for auditing.
• Removing the IAM role removes their login access.

• compute.osLogin: regular OS login access
• compute.osAdminLogin: OS login with administrative privileges, typically including sudo

text 
enable-oslogin=TRUE

Box 1: No

• configure_azure_monitor() configures Azure Monitor and the OpenTelemetry pipeline.
• LangChain requires its own instrumentation or callback tracer, such as AzureAIOpenTelemetryTracer.
• You must pass the tracer to the LangChain execution configuration so chains, tools, prompts, and model calls become spans in Application Insights.

Box 2: Yes

text 
OTEL_SERVICE_NAME=customer-agent 
OTEL_SERVICE_NAME=invoice-agent

DAX 
Previous Year Sales = 
CALCULATE( 
    [Total Sales], 
    PARALLELPERIOD('DATE'[Month], -1, YEAR) 
)

• CALCULATE evaluates [Total Sales] after modifying the filter context.
• PARALLELPERIOD shifts the current date period. -1 moves it backward, and YEAR specifies a one-year shift.
• 'DATE'[Month] is the date column used for the time calculation. In a proper date table, this should be a real date column or an appropriate calendar column—not merely text such as "January".

• Due diligence: Investigating and evaluating a third party before entering a business relationship.
• Attestation: An independent auditor’s examination and formal report—such as the SOC 2 report itself. The analyst is reviewing the attestation, not performing it.
• Internal audit: Examines the organization’s own processes and controls.
• Penetration testing: Actively tests systems for exploitable vulnerabilities.

• Preparing and managing datasets
• Building and training machine-learning models
• Testing and evaluating models
• Deploying models as endpoints for predictions

• Root path cost
• Port priority
• Port ID

text 
interface GigabitEthernet0/1 
 spanning-tree vlan 10 port-priority 64

• Interface number: Helps form the port ID but is not normally configured as the preference value.
• VLAN priority: Not used to select a preferred forwarding interface.
• Hello time: Controls how often BPDUs are sent; it does not select a preferred port.

• A: CEO-sponsored formal policy — demonstrates executive support, but sponsorship at one point in time may not sustain commitment.
• B: Security awareness training — primarily improves employee behavior, not senior management commitment.
• C: Periodic alignment reviews — ongoing reviews show that security remains relevant, supports business priorities, and uses resources appropriately.
• D: Senior management signoff on the strategy — useful for governance and accountability, but signoff is generally a one-time event.

• WPA3: Secures wireless traffic but does not normally verify whether someone has paid.
• PSK: A shared Wi-Fi password gives access to anyone who knows it; it cannot easily validate individual receipts.
• IEEE 802.1X: Provides user/device authentication, commonly through an authentication server, but is excessive and not designed specifically for a simple receipt-based guest-access workflow.

• Identity: User-assigned managed identity

• Authorization: Azure RBAC

• A. PsFile — Correct: Displays files opened remotely on a system. This reveals shared resources and user activity.
• B. UserInfo — Correct: Gathers information about user accounts, supporting account and privilege enumeration.
• C. PsPasswd — Answer key lists it: This is questionable. PsPasswd is primarily used to change passwords remotely, not enumerate information. The key may classify it broadly as a Windows account-management tool, but it is not a strong enumeration example.
• D. WinSSLMiM — Incorrect: Used for SSL man-in-the-middle activity, not enumeration.

• Wireshark captures packets and allows you to examine their contents, potentially revealing usernames and passwords sent without encryption.
• Netcat creates network connections and can send or receive data, but it is not primarily a packet-analysis tool.
• Nessus scans for known vulnerabilities and configuration issues; it generally would not prove that credentials are visible in transit.
• Nmap performs network discovery and port/service scanning, not detailed packet-content inspection.

• Accept: Acknowledge the risk and take no additional action.
• Transfer: Shift the risk or financial impact to another party, such as through insurance or a contract.
• Mitigate: Reduce the likelihood or impact using controls, such as firewalls, training, or backups.
• Avoid: Eliminate the activity that creates the risk.

• Crowd-sourced evaluation: Requires recruiting and managing many evaluators.
• Human-worker evaluation: Provides human judgment but requires coordination, labeling instructions, and quality control.
• RLHF: Is a training and alignment process, not simply a low-overhead model comparison method.
• Automatic model evaluation: Quickly evaluates outputs using predefined toxicity or quality metrics with minimal manual effort.

• Asynchronous inference is designed for applications that submit individual inference requests, place them in a queue, and retrieve results later. It is useful when requests may take a long time or when traffic is intermittent.
• Batch transform is designed specifically to run inference over an entire large dataset stored in Amazon S3 and produce an output dataset.

• Large dataset/offline processing ? Batch transform
• Large or slow individual requests processed later ? Asynchronous inference

• netstat -tuln can show listening TCP ports.
• If the web service is not listening, or is bound only to 127.0.0.1, remote browsers cannot connect even though the server responds to ping.

• traceroute: Tests the path to the server, but ping already confirms basic reachability.
• tcpdump: Useful for detailed packet analysis, but it is more appropriate after checking whether a service is listening.
• arp: Helps troubleshoot local IP-to-MAC resolution, not an application-layer web connection.

• ?? ?? ?? ??
• ?? NFS ?? SMB ?? ??
• ?? ???? ???
• ?? IP ??? ???? ?? ???
• ??? ?? ???? ?????

• B: ????? NFS ? Snowball Edge? ?? ?? ? AWS? ?? ? S3
• D: ????? NFS ? ????/Direct Connect ? File Gateway ? S3

• A hypervisor allows multiple virtual machines to run on one physical server.
• Each VM can have its own operating system, so the customer can consolidate Windows, Linux, and Unix servers.
• A Type 2 hypervisor runs on a host operating system and can be deployed on local hardware, helping address data privacy and avoiding ongoing public-cloud charges.

• Containers share the host operating system’s kernel.
• They are well suited to consolidating applications, but they generally cannot independently run a mix of unrelated kernels such as Windows, Linux, and Unix on one host.

• Label applied does not automatically exclude a file from Copilot.
• Encryption/usage rights determine whether Copilot can access the file’s content.
• If the user has no permission under Label2, File3 should be unavailable to Copilot for that user.
• File3’s label might be inherited by derived content only if File3 was actually processed and contributed by an authorized user; it should not be treated as contributing merely because it exists in the source set.

• Matching hashes: The software likely has not changed.
• Different hashes: The software was modified or corrupted.

• Encryption: Protects confidentiality by making data unreadable; it does not primarily detect changes.
• Baselines: Define an approved system or configuration state, but they are not the specific integrity mechanism here.
• Tokenization: Replaces sensitive data with substitute values.