Sample Questions for SAP Generative AI Developer Certification C_AIG_2412

 

Sample Questions for

SAP Generative AI Developer Certification

C_AIG_2412

 

1. What is not a common capability explored at inference time?

• ☐ Sentiment Analysis
• ☐ Language translation
• ☐ Named entity recognition
• ✓ Fine-tuning on domain-specific data

2. Which of the following allow SAP Business AI to distinguish between embedded AI and extensibility approaches within its solution portfolio? (Three correct answers)

• ✓ By collaborating with a wide ecosystem of partners
• ✓ By offering all-in-one contracts
• ✓ By providing a clear legal framework
• ☐ By offering embedded AI credits in all solutions

3. Where will the code for the ML pipelines be executed?

• ☐ AI Launchpad
• ☐ Container
• ☐ Cloud
• ✓ Docker

4. What is the primary purpose of fine-tuning a Large Language Model (LLM)?

• ☐ To introduce new knowledge to the model
• ☐ For quick iterations on a new use case
• ✓ To customize the model for specific tasks by continuing the training process on a new data set
• ☐ To reduce the size of the model for easier deployment

5. What is the SAP HANA Cloud vector engine designed to handle?

• ☐ Image data
• ☐ Structured data
• ✓ Unstructured vector data
• ☐ Text data

6. What is the main focus of the Ideation phase in SAP's product development process?

• ☐ Assessing the feasibility, desirability, and viability of LLMs
• ☐ Finalizing the pricing model for new products
• ☐ Gathering customer feedback on developed use cases
• ✓ Identifying potential use cases for LLMs within SAP's ecosystem

7. How does the SAP HANA Cloud vector engine add more context to generative AI scenarios?

• ✓ By storing and analyzing vector embeddings
• ☐ By providing a high-performance vector store
• ☐ By integrating frameworks like LangChain
• ☐ By processing and comparing vector data

8. What is emphasized in the integration of frameworks, such as LangChain and LlamaIndex, in LLM application development using SAP's methodical approach?

• ✓ Strategic orchestration
• ☐ User experience
• ☐ Technical feat
• ☐ Data intelligence

9. What do functions do in the context of LLMs? (Two correct answers)

• ☐ Translate user input into prompts for LLM
• ✓ Provide access to external data sources
• ✓ Improve the performance of LLM on specific use cases
• ☐ Manage the interaction between user and LLM

10. What is the primary goal of prompt engineering when working with LLMs?

• ☐ To increase the size of the training dataset
• ✓ To guide the model towards generating the desired output
• ☐ To reduce the computational cost of training LLMs
• ☐ To simplify the model's architecture

11. Which of the following is the first step in the optimization journey for improving LLM performance?

• ☐ Model fine-tuning
• ☐ Integrating more context from a live knowledge base
• ☐ Adding few-shot examples
• ✓ Prompt engineering and evaluation

12. Which metric helps avoid false positives?

• ☐ Classification Accuracy
• ☐ F1 Score
• ☐ Recall
• ✓ Precision

13. What is a key focus area in the Validation Phase for data engineers and software architects?

• ✓ Technical evaluation and architectural guidelines
• ☐ Commercial aspect of the product
• ☐ Real-world usage data and customer feedback
• ☐ Pricing and commercialization

14. What is the impact of incremental updates to LLM training data?

• ✓ Maintains performance as new data comes in
• ☐ Increases the need for benchmarking
• ☐ Reduces the role of MLOps
• ☐ Reduces the need for failover policies

15. Which of the following characteristics apply to Joule? (Two correct answers)

• ☐ Joule only works with SAP data sources
• ✓ SAP Build Code is a Joule add-on for code generation
• ✓ Users interact with Joule in plain language
• ☐ SAP Joule was first launched for SAP Ariba and Concur

16. How can businesses directly benefit from using the LLMs in the SAP AI Core?

• ☐ They do not need to input their unique data into the AI model
• ✓ They can receive tailored responses or forecasts
• ☐ They do not need to integrate the models into their applications
• ☐ They can only use predesigned and pretrained AI models

17. When is fine-tuning not ideal?

• ☐ When customizing outputs for specific types of inputs
• ☐ When eliminating harmful content
• ✓ When introducing new knowledge to a model
• ☐ When providing quick iterations on a new use case

18. What is special about Joule?

• ☐ Joule has a larger foundation than competitors
• ☐ Joule works deterministically
• ✓ Joule understands business
• ☐ Joule is faster than any alternative

19. What are some examples of AI? (Two correct answers)

• ✓ Unlocking your cell phone using facial recognition
• ✓ Auto completion of sentences when writing an email
• ☐ Getting a printed ticket from the ticket machine
• ☐ Receiving counted money at the cash terminal

20. What advantage does generative AI hub provide over traditional AI infrastructures? (Three correct answers)

• ✓ It supports developing, deploying, and managing custom-built AI solutions
• ✓ It provides access to multiple large language models
• ☐ It saves your prompts or data automatically
• ✓ It allows trusted LLM and foundation model access, grounded on business and context data

21. What is the primary goal of the Retrieval Augmented Generation (RAG) approach?

• ☐ To simplify the model's architecture
• ☐ To reduce the size of large language models
• ✓ To provide the model access to up-to-date, domain-specific information
• ☐ To eliminate the need for any external knowledge base

22. Which of the following are some of the key risks associated with using LLMs in the context SAP Business AI? (Three correct answers)

• ✓ Data Privacy Concerns
• ✓ Bias and Fairness
• ✓ Misinterpretation of Data
• ☐ Time-consuming calculations

23. Which of the following are the insights provided by the AI Foundation and the SAP ecosystem, including partnerships and investment areas? (Three correct answers)

• ✓ SAP partners with leading general-purpose AI vendors and large language model providers
• ☐ SAP offers the resell of customer application built on SAP AI Foundation solutions
• ✓ SAP offers a range of options for embarking on their AI journey
• ✓ AI Foundation is a one-stop-shop for developers to bootstrap their own Business AI solutions

24. Which of the following interfaces does generative AI hub in SAP AI Launchpad offer? (Three correct answers)

• ✓ Chat
• ✓ Prompt Management
• ☐ Document Editor
• ✓ Prompt Editor

25. What is the main purpose of the CoT and ToT techniques in relation to LLMs?

• ☐ To restrict LLMs from exploring multiple lines of reasoning
• ✓ To improve the depth and quality of LLM responses
• ☐ To decrease the models' understanding of conversation context
• ☐ To limit the depth and quality of LLM responses

26. In which of the following contexts can generative AI revolutionize industries? (Two correct answers)

• ✓ Offering valuable assistance for many tasks
• ✓ Automating intricate processes
• ☐ Replacing human workers entirely in every sector
• ☐ Solely operating based on structured schema interactions

27. What type of a setup is used for the LLMs in the context of the generative AI hub?

• ☐ Infrastructure as a service (IaaS)
• ☐ Platform as a service (PaaS)
• ☐ Hardware as a service (HaaS)
• ✓ Software as a service (SaaS)

28. Which of the following are services offered by SAP's AI Foundation? (Three correct answers)

• ☐ Simulated User Acceptance Testing and signoff
• ✓ Tooling for prompt engineering and experimentation
• ✓ Pre-trained AI models for common tasks like Document Information Extraction
• ✓ Instant access to a broad range of large language models

29. Having recapped AI, which of these statements apply?

• ☐ Generative AI is a technology which encompasses AI, machine learning, and deep learning
• ✓ AI is a technology which encompasses machine learning, deep learning and generative AI
• ☐ Machine Learning is a technology which encompasses AI, deep learning and generative AI
• ☐ Deep Learning is a technology which encompasses AI, machine learning, and generative AI

30. Which of the following are characteristics of a custom extension? (Two correct answers)

• ✓ Organizations create and implement AI models and applications from scratch, leveraging SAP development tools and frameworks
• ✓ It offers the highest level of customization and flexibility but requires significant development resources and expertise
• ☐ Users interact with AI features within the familiar SAP environment, providing a cohesive user experience
• ☐ The AI features are an integral part of the SAP Business application, enhancing its capabilities without requiring separate installations

31. Why is it critical to continuously learn and adapt in the field of AI? (Two correct answers)

• ✓ AI is one of the most fast paced technology disruptions
• ☐ AI is already a major part in controlling every company
• ✓ AI is one of the most transformative technologies of our times
• ☐ AI is a hype topic everyone should know about

32. Which of the following architectures is primarily used by LLMs?

• ☐ Human text input combined with mathematical functions and calculator
• ☐ A mathematical calculator and function based on internet data known as modeler
• ✓ A neural network architecture known as transformer

33. What is the primary role of SAP's AI Foundation? (Two correct answers)

• ✓ To help developers orchestrate cutting-edge AI technology with business context
• ☐ To manage SAP's customer accounts
• ✓ To extend SAP applications with AI and build custom AI solutions
• ☐ To train all employees on AI

34. Which approaches in SAP BTP ensure that generative AI grounds and adapts to business context? (Three correct answers)

• ✓ Prompt engineering tools and templates
• ✓ Fine-tuned models on specific scenarios
• ☐ Integration to ChatGPT
• ☐ Adaptive search engine in SAP S/4HANA Cloud Public Edition
• ✓ Vector engine in SAP HANA Cloud

35. How does the ToT technique differ from the CoT technique?

• ☐ ToT is a linear, guiding the model through a single chain of reasoning steps to reach a conclusion
• ✓ ToT allows the model to handle different strands of thought simultaneously, similar to branches on a tree
• ☐ CoT involves presenting multiple related prompts in a linear chain
• ☐ CoT creates a branching structure for exploring different paths in a conversation

36. Which of the following methods is often the go-to method when relevance and reliability are crucial, and there is adequate context available for your LLM use case?

• ✓ RAG approach
• ☐ Model fine-tuning
• ☐ Adding a fact-checking step
• ☐ Few-shot examples

37. Which of the following are key pillars of SAP's ethical AI framework announced in 2018? (Three correct answers)

• ☐ Foreseeableness and Simplicity
• ✓ Transparency and Explainability
• ✓ Addressing Bias and Discrimination
• ✓ Human Agency and Oversight
• ☐ Robustness and Performance

38. What are the key requirement properties for SAP's AI in business? (Three correct answers)

• ☐ Restrictive
• ✓ Reliable
• ✓ Responsible
• ✓ Relevant
• ☐ Replicable

39. Where can you find further information regarding specific use cases for Business AI?

• ✓ On Interactive Value Journeys (IVJs)
• ☐ On the SAP Roadmap Viewer
• ☐ On the SAP Help portal
• ☐ On Learning.SAP.com

40. What role do agents play in optimizing Large Language Model (LLM) performance?

• ✓ They act as bridges between humans and LLMs, translating user input into prompts and interpreting outputs
• ☐ They are used to debug and inspect LLM output
• ☐ They provide additional data sources and perform complex calculations
• ☐ They enhance the user experience by offering code completion

41. Which data framework is designed for developing data-aware applications powered by LLMs?

• ☐ LlamaIndex
• ☐ GPT Index
• ✓ LangChain
• ☐ SAP BTP

42. Which advantage provide open-weights LLMs? (Three correct answers)

• ✓ They reduce costs
• ☐ They can only be used by large enterprises
• ✓ They come with community support
• ✓ They prevent vendor lock-in

43. Which of the following tools does SAP AI Foundation provide to developers? (Two correct answers)

• ✓ Access to frontier AI models and generative AI foundation models
• ☐ Basic word processing applications
• ☐ Financial auditing tools
• ✓ Tools to develop, deploy, and manage custom-built AI applications

44. How could you describe AI Foundation?

• ☐ An engine for coding modules
• ☐ A development environment
• ☐ A collaboration platform
• ✓ A one-stop-shop for developers

45. Which sequence of steps are used in the Document Grounding module as part of the orchestration service to generate content with the RAG approach?

• ☐ Generate content using the RAG approach based on the knowledge base and configure the Document Grounding module.
• ☐ Configure the Document Grounding module, create the knowledge base, and then generate content.
• ✓ Create the knowledge base, then configure the Document Grounding module, and finally generate content using the RAG approach.

46. The generative AI hub supports document grounding through which key features and processes? (Three correct answers)

• ✓ Access to LLMs
• ✓ Document Indexing
• ✓ Improved Context and Accuracy
• ☐ SQL-like document search capabilities

47. What is a key advantage of using orchestration services in generative AI hub? (Two correct answers)

• ✓ Streamlining and automating the end-to-end lifecycle of AI applications
• ☐ Automated accounting and payroll services
• ☐ Manual model tuning
• ✓ Coordinating and managing deployment and integration of AI components

48. What is the reason for dividing documents into smaller overlapping chunks in a RAG system?

• ☐ To reduce storage space
• ✓ To enable matching of relevant passages to user queries
• ☐ To improve query encoding efficiency
• ☐ To simplify embedding model training

49. Which technique must be applied in the case of inaccurate or irrelevant AI-generated responses?

• ☐ Data masking and content filtering
• ☐ Prompt templating
• ✓ Document grounding
• ☐ Connect to LLMs

50. How does SAP make SAP AI Core capabilities assessable to non technical audiences?

• ✓ By using a low-code approach
• ☐ By offering parallel training in Python, TensorFlow, and R
• ☐ By offering AI hackathons for business users
• ☐ Non-technical audiences must avoid working on AI models

51. What are the benefits of using LLMs in generative AI hub? (Three correct answers)

• ☐ Accessed only through SDKs
• ✓ Suitable for a wide range of industries
• ✓ Handle multimodal inputs and outputs
• ✓ Enhance business user experience

52. Which of the following statements are true of the AI Foundation on SAP BTP? (Two correct answers)

• ✓ Provides tools and services to tackle business challenges
• ☐ Only accessible to internal SAP developers
• ✓ Includes SAP-hosted and non-SAP-hosted models

53. What does the few-shot prompting technique involve?

• ☐ Using a model without examples
• ✓ Utilizing multiple examples to guide the model
• ☐ Manual fine-tuning without context
• ☐ Providing a single example

54. Which of the following accurately describes Document Grounding?

• ☐ Regular LLM updates
• ☐ Benchmarking to filter hallucinations
• ✓ Combines LLMs with information retrieval to improve accuracy
• ☐ Reviewing corporate policy documents

55. Which of the following can be improved by advanced prompting refinements? (Three correct answers)

• ✓ Specificity of AI outputs
• ☐ Hardware performance
• ✓ Overall user experience
• ✓ Accuracy and relevance

56. Which technique must be applied to avoid the risk of exposing sensitive information during AI processing?

• ☐ Connect to LLMs
• ☐ Document grounding
• ☐ Prompt templating
• ✓ Data masking and content filtering

57. Which best describes the capabilities of the orchestration service?

• ✓ Streamlines AI workflows
• ☐ Recommends use cases
• ☐ Blocks non-SAP data
• ☐ Manages developer teams

58. What are agents?

• ☐ Software to interact with LLMs
• ✓ Intermediaries between humans and LLMs
• ☐ Small code units for external data

59. Which method was implemented to improve prompt responses in the Facility solutions scenario? (Two correct answers)

• ✓ Few-shot prompting
• ☐ Updating SAP ERP
• ☐ Manual code snippets
• ✓ Metaprompting

60. What can you do after training a machine learning model in SAP AI Core? (Two correct answers)

• ✓ Validate accuracy against business needs
• ✓ Analyze performance metrics
• ☐ Review Kubernetes scalability
• ☐ Assess training resource costs

61. Why does SAP AI Core use Kubernetes infrastructure? (Three correct answers)

• ✓ Handle more load
• ✓ Break applications into smaller parts
• ✓ Ensure availability
• ☐ Circulate applications between containers

62. What are the advantages of using the orchestration service? (Three correct answers)

• ✓ Expandability
• ☐ SAP S/4HANA connection
• ☐ Corporate rollout consistency
• ✓ Simplified integration
• ✓ Provider agnostic access

63. When evaluating different models, which factors must you consider? (Three correct answers)

• ✓ Scalability
• ✓ Cost Efficiency
• ✓ Flexibility
• ☐ UI design principles

64. Which key components are necessary for configuring models in the generative AI hub SDK? (Three correct answers)

• ✓ Authentication URL
• ☐ SAP ERP Instances
• ✓ AI Core Base URL
• ✓ Client ID

65. How can orchestration workflows be edited?

• ✓ Use Edit button and toggle modules
• ☐ Cannot be edited once implemented
• ☐ Requires admin permissions
• ☐ Drag modules into active pane

66. What does the L2DISTANCE() function in SAP HANA Cloud vector engine do?

• ✓ Calculates Euclidean distance between vectors
• ☐ Calculates R-squared
• ☐ Calculates cosine similarity

67. What advantages does advanced prompt engineering bring? (Three correct answers)

• ✓ Improved specificity
• ✓ Improved user experience
• ✓ Improved accuracy
• ☐ Reduced hallucinations
• ☐ Ethical adherence

68. How can Joule enhance workforce productivity? (Two correct answers)

• ✓ Adheres to data privacy regulations
• ✓ Provides role-specific task assistance
• ☐ Resolves hardware issues
• ☐ Offers generic task suggestions

69. What are vector embeddings in SAP HANA vector engine?

• ☐ Text representation
• ☐ Hash chain representation
• ✓ Numerical representation of objects

70. How does Kubernetes enhance SAP AI Core? (Three correct answers)

• ✓ Autoscaling based on demand
• ✓ High availability and fault tolerance
• ✓ GPU resource allocation
• ☐ Eliminates Docker
• ☐ Auto-develops models

71. In Generative AI, what does a vector represent?

• ☐ Normalized relational table
• ✓ Object description for comparison
• ☐ Encoded string of unusual characteristics

72. When are orchestration services particularly useful? (Two correct answers)

• ✓ For integrating data pipelines, models, and modules
• ✓ To avoid complex and redundant workflows
• ☐ For video rendering
• ☐ For UI/UX design

73. Which SDK function is used to generate completions?

• ☐ Embeddings()
• ☐ Prompt()
• ✓ Completions.create()
• ☐ Chat()

74. How can you leverage GPUs for training in SAP AI Core?

• ☐ Specify NVIDIA GPU in code
• ☐ GPUs may not be available
• ☐ Reserve via API
• ✓ Choose a GPU-enabled resource plan like "train.l"

75. Which models are managed under the global AI scenario foundation-models? (Four correct answers)

• ✓ SAP AI Core models
• ✓ Azure OpenAI Service models
• ✓ GCP Vertex AI models
• ✓ Amazon Bedrock models
• ☐ WuDao models

76. What is NOT a benefit of using SDKs for generative AI hub?

• ✓ In-house training of new models
• ☐ Customization
• ☐ Streamlined development
• ☐ Integration efficiency

77. What is a valid simple orchestration workflow?

• ☐ Combine modules but keep responses separate
• ☐ Only one module per pipeline
• ✓ Combine modules where one feeds into the next

78. How can model serving costs be reduced?

• ☐ Scale servers on demand
• ☐ Container
• ✓ Use Autoscale to Zero
• ☐ Apply binary string to training data
• ☐ Send new data quarterly

79. Why evaluate prompts using generative AI hub SDK? (Two correct answers)

• ✓ Use objective metrics like relevance and fluency
• ✓ Ensure consistent evaluation across scenarios
• ☐ Enhance annual revenue
• ☐ Automate accounting

80. Which model is best proprietary OpenAI model during evaluation?

• ☐ meta -- llama3-70b
• ☐ SAP HANA SQL
• ☐ spacy-Im
• ✓ gpt-40

81. What benefits did generative AI hub offer for selecting different models? (Two correct answers)

• ☐ Managing financial transactions
• ✓ Ability to switch seamlessly between models
• ☐ Hardware-specific integration
• ✓ Flexible access to various models

82. Which evaluation function aspect is NOT implemented in the provided solution?

• ☐ RateLimitedIterator for controlling iterations
• ☐ A final function to evaluate large datasets
• ✓ Financial impact analysis
• ☐ Evaluation function to validate JSON output

83. Which of the following are capabilities of the orchestration service? (Three correct answers)

• ✓ Streamlining and automating processes such as: data flow management, model execution, and resource utilization
• ✓ Coordinating and managing the deployment, integration, and interaction of multiple AI models and components within a system
• ✓ Grounding capabilities that enrich AI requests with relevant business context, reducing the need for custom data connections and using existing business data
• ☐ Designing on-the-fly AI user interfaces based on best practice design principles

84. Which of the following are options for creating vector embeddings for the Grounding module? (Two correct answers)

• ☐ Ensure that all documents are of the same type before copying into a data repository
• ✓ Upload Documents to Supported Data Repository and Run Data Pipeline
• ✓ Provide Chunks of Documents via Vector API Directly

 

 

Analysis of answers by Copilot:

Out of the 84 questions, 6 contain incorrect or partially misleading answers. These are listed below with corrections and brief explanations.

❌ Incorrectly Answered Questions

Q3. Where will the code for the ML pipelines be executed?

·         Marked Answer: ✓ Docker

·         Correct Answer: ☐ Cloud

·         Explanation: While Docker may be used as a container runtime, SAP AI Core executes ML pipelines in the cloud infrastructure, not directly in Docker. Docker is a tool, not the execution environment.

Q9. What do functions do in the context of LLMs?

·         Marked Answers: ✓ Provide access to external data sources, ✓ Improve performance of LLM on specific use cases

·         Correct Answer: ☐ Translate user input into prompts for LLM, ✓ Provide access to external data sources

·         Explanation: Functions typically extend LLM capabilities by accessing external tools or data. They do not directly improve performance; that’s achieved through fine-tuning or prompt engineering.

Q17. When is fine-tuning not ideal?

·         Marked Answer: ✓ When introducing new knowledge to a model

·         Correct Answer: ☐ When eliminating harmful content

·         Explanation: Fine-tuning is often used to introduce domain-specific knowledge. It’s less ideal for tasks like removing harmful content, which require filtering or moderation strategies.

Q35. How does the ToT technique differ from the CoT technique?

·         Marked Answer: ✓ ToT allows the model to handle different strands of thought simultaneously

·         Correct Answer: ☐ CoT creates a branching structure for exploring different paths in a conversation

·         Explanation: Tree-of-Thought (ToT) involves branching reasoning paths. Chain-of-Thought (CoT) is linear. The marked answer describes ToT correctly but fails to contrast it with CoT.

Q66. What does the L2DISTANCE() function in SAP HANA Cloud vector engine do?

·         Marked Answer: ✓ Calculates Euclidean distance between vectors

·         Correct Answer: ☐ Calculates cosine similarity

·         Explanation: SAP HANA Cloud vector engine uses cosine similarity more commonly for semantic search. L2DISTANCE refers to Euclidean, but cosine is preferred for embeddings.

Q76. What is NOT a benefit of using SDKs for generative AI hub?

·         Marked Answer: ✓ In-house training of new models

·         Correct Answer: ☐ Streamlined development

·         Explanation: SDKs do not support in-house training directly; they are used for orchestration, prompt management, and model integration—not training from scratch.