An AI team is planning two separate projects. The architect needs to provision the appropriate infrastructure for each.
| | Project A | Project B|
| -- | | - |
| Goal | Build a novel image recognition model from scratch.
| Adapt an existing, pre- trained LLM to understand company-specific jargon. |
| Input Data | 10 million new, unlabeled images. | A 50 GB text corpus of internal documents. |
| Required Compute | Very High (Weeks of multi-GPU training) | Moderate (Hours of single-GPU training) | Which two statements accurately describe the infrastructure requirements for these projects?
(Choose two.)

• A. Project B is a fine-tuning task, which leverages an existing model and requires less data and compute resources than building a model from scratch.
• B. Project A is a model building task and requires a significantly larger and more powerful compute and storage infrastructure than Project B.
• C. Both projects should use NetApp StorageGRID as the primary storage for training to ensure low latency.
• D. Both projects are examples of fine-tuning and have similar infrastructure needs.
• E. Project A requires only a large amount of data, while Project B requires only a large amount of code.

The company's compliance officer mandates that all raw simulation data in the StorageGRID data lake must be protected from modification or deletion for a period of seven years. Which two StorageGRID features should be implemented to meet this strict, long-term data immutability requirement? (Choose 2.)

• A. Enable Autonomous Ransomware Protection (ARP) on the StorageGRID system.
• B. Configure an Information Lifecycle Management (ILM) rule that applies to the bucket and sets a retention period of seven years.
• C. Create a new S3 bucket with Object Lock enabled in Compliance mode.
• D. Use NetApp SnapMirror to replicate the data to a second StorageGRID system.
• E. Encrypt all objects in the bucket using server-side encryption (SSE).

An architect is designing a cost-optimized storage solution for a large AI environment that has two distinct data temperature profiles:
1. Hot Data (200 TB): Actively used training sets and models requiring all-flash performance.
2. Cold Data (2 PB): Inactive, archived datasets and old model versions that must remain online but are infrequently accessed.
The solution must automatically manage data placement based on activity, without requiring manual intervention or changes to file paths. The environment consists of an on-premises data center and a public cloud account.
Which combination of NetApp technologies provides the most complete and cost-effective solution? (Select all that apply.)

• A. Use NetApp SnapMirror to create a full copy of the 2 PB of cold data on the AFF A-Series for faster access if needed.
• B. Use NetApp FabricPool to automatically and transparently tier cold data blocks from the AFF A- Series to the StorageGRID system.
• C. Deploy NetApp StorageGRID on-premises to serve as a cost-effective capacity tier for the cold data.
• D. Use NetApp XCP to manually script the daily movement of cold files from the AFF system to StorageGRID.
• E. Deploy a single, large Cloud Volumes ONTAP instance to hold both the hot and cold data tiers.
• F. Deploy a NetApp AFF A-Series system on-premises to serve as the high-performance tier for the hot data.

What is the primary architectural advantage of using a NetApp AIPod with NVIDIA DGX servers for the AI training cluster, as described in the scenario?

• A. It exclusively uses object storage, which simplifies access for data scientists using S3-native tools.
• B. It is a reference architecture that is pre-validated by NetApp and NVIDIA to eliminate design complexity and ensure predictable performance for AI workloads.
• C. It prioritizes CPU performance over GPU performance for traditional machine learning algorithms.
• D. It is designed for small-scale, departmental AI projects and cannot be scaled.