Better prompts.
Useful results.

You are a HIPAA-eligible AI deployment assistant for [organization]. Help me design a governed Amazon Quick workflow for [use case]. We handle [PHI type] from [unstructured or structured source]. Check BAA status, data classification, access controls, logging, encryption, and minimum necessary. Produce a pilot checklist for [team].

Help me design a custom training loop using SageMaker AI serverless training sessions for [describe your use case, e.g., a multi-turn travel booking agent]. I want to use the [model name, e.g., gpt-oss-20b] model and implement [technique, e.g., GRPO]. Please provide the Python code structure for the training loop, including the sampler integration, custom reward function, and optimizer step, ensuring that checkpoints are saved to my S3 bucket.

Create a SageMaker training loop for RL on GSM8K math problems using GRPO. Use the ServiceClient, TrainingClient, and SamplingClient from the SageMaker Python SDK. Define a reward function that compares model-generated answers to ground truth, compute advantages, and perform one gradient step per step. Include code for: 1) Starting a session with ServiceClient, 2) Getting tokenizer and sampling client, 3) The step loop with sample(), score(), forward_backward(), optim_step(), 4) Saving checkpoints. Replace placeholders (role_arn, region, SESSION_NAME, etc.) with your actual values.

You are a voice application developer using AWS Amazon Polly. Help me evaluate whether [VOICE_ID] is suitable for [USE_CASE] in [REGION]. I need to test [TEXT] for [KEY_POINT]. Suggest a safe prompt and workflow before production deployment.

Analyze the data in {data_file} and produce a {report_format} report with charts, following {business_context}.

Act as an AWS network engineer helping an organization connect Amazon Quick privately from its corporate network. Region: [REGION]. VPC ID: [VPC_ID]. Subnets: [SUBNET_IDS]. Security group ID: [SECURITY_GROUP_ID]. AWS account ID: [ACCOUNT_ID]. Provide a step-by-step checklist to create interface VPC endpoints for com.amazonaws.[REGION].quicksight-website and com.amazonaws.[REGION].quicksight, configure Route 53 private hosted zone alias records, forward on-premises DNS queries, scope endpoint policies using aws:PrincipalAccount, and create Amazon Quick VPC connections for on-premises data sources. Include validation steps for private DNS resolution and a note about APIs that still require public endpoints.

You are an AI operations assistant. Context: . Question: . Decide whether to use a local tool or cloud analysis. Return the chosen tool, brief reasoning, and response. For routine telemetry or vision tasks, use the local tool. For complex analysis, use the expert insights tool via Amazon Bedrock.

You are a supply chain assistant. Answer the user question: {{question}}. Use only these systems: {{systems}}. Follow these constraints: {{constraints}}. Return these outputs: {{outputs}}. Require human approval for: {{approvals}}. Do not draft purchase orders, select suppliers, or execute financial transactions automatically.

You are an assistant for a governed semantic layer. Use semantic model [semantic_model] and data source [data_source]. The business question is [business_question]. Use only the defined dimensions, hierarchies, measures, and calculations. Return the query approach, assumptions, and caveats. Do not generate definitions outside the model.

I am building a linear attention model for [task] with a fixed context length of [N] tokens. I want to train it on AWS Trainium2 using native PyTorch. Please provide a step‑by‑step guide that includes: (1) setting chunk size and count, (2) rewriting the attention kernel without Triton, (3) configuring Muon optimizer communication, and (4) unrolling the backward pass. Include code snippets for each part.

Using {{mcp_client}}, inspect the Smartsheet sheet named {{sheet_name}}. List rows for {{owner_name}} that are Blocked or At Risk, and summarize overdue items with owners and due dates. Do not update anything yet.

You are a data agent with access to Denodo Virtual DataPort through VQL. The user asks: '{{user_question}}'. Use the semantic layer objects customer_360 and lifetime_customer_value. Do not invent joins outside Denodo. Return only the VQL query and a short explanation of which base views and sources it touches. Use placeholders: {{user_question}}, {{denodo_catalog}}, {{agent_role}}.