準備好獲得 Databricks 認證：生成式 AI 工程師助理.pdf

《準備好獲得 Databricks 認證：生成式 AI 工程師助理.pdf》由會員分享，可在線閱讀，更多相關《準備好獲得 Databricks 認證：生成式 AI 工程師助理.pdf（17頁珍藏版）》請在三個皮匠報告上搜索。

1、2024 Databricks Inc.All rights reservedJames Kantor&Corey AbshireJames Kantor&Corey Abshire1GET READY TO BE GET READY TO BE DATABRICKS CERTIFIED:DATABRICKS CERTIFIED:Generative AI Generative AI Engineer AssociateEngineer Associate2024 Databricks Inc.All rights reservedJames KantorSr.Certification De

2、veloper2YOUR SPEAKERSYOUR SPEAKERSCorey AbshireSr.Specialist Solutions Architect2024 Databricks Inc.All rights reserved2024 Databricks Inc.All rights reserved3DATABRICKS DATABRICKS CERTIFIEDCERTIFIEDGENERATIVE AIGENERATIVE AIENGINEER ENGINEER ASSOCIATEASSOCIATE2024 Databricks Inc.All rights reserved

3、4CERTIFICATION BACKGROUNDCERTIFICATION BACKGROUND Get certified in todays GenAI and DatabricksExam=cutting edge LLM/Chained/Agents/RAG solutions PLUS Databricks A new role path for Databricks learning:GenAIML path is separate and not a prerequisite One exam to certifyStandard 2 year validity2024 Dat

4、abricks Inc.All rights reserved5EXAM DETAILSEXAM DETAILS 45 questions x 90 minutes Scenario-based questionsMultiple choice or multiple selection$200/attempthttps:/ Databricks Inc.All rights reserved6WHAT IS THE SCOPE OF THIS EXAM?WHAT IS THE SCOPE OF THIS EXAM?In-Scope Problem Decomposition Selectin

5、g Models,Tools,and ApproachesPrompt engineering,current APIs,evaluation and monitoring Databricks TechnologyVector Search,Model Serving,MLflow,Unity CatalogOut-of-ScopeFine tuning,Pre-training from scratch,Continued pre-training,Model Architecture2024 Databricks Inc.All rights reserved7THE EXAM TARG

6、ET(MQC)THE EXAM TARGET(MQC)MQC is the Minimally Qualified CandidateIdeal candidate for the exam MQC can design and implement LLM-enabled solutions using Databricks MQC has consumed all learning and has six months hands-on experience2024 Databricks Inc.All rights reserved8THE EXAM GUIDETHE EXAM GUIDE

7、2024 Databricks Inc.All rights reserved9MAJOR SKILL AREASMAJOR SKILL AREAS Problem Decomposition Selecting Models,Tools,and ApproachesPrompt engineering,LangChain,Hugging Face,current APIs,LLM Evaluation Databricks TechnologyVector SearchModel ServingMLflowUnity Catalog2024 Databricks Inc.All rights

8、 reserved10EXAM SECTIONSEXAM SECTIONS1.Design Applications2.Data Preparation3.Application Development4.Assembling and Deploying Applications5.Governance6.Evaluation and Monitoring2024 Databricks Inc.All rights reservedPrompt designModel selectionChainingFunction calling(tools)Multi-stage reasoningCh

9、unking documentsData preparationContent pipelinesData sourcesRetrieval evaluationData retrievalImplementing guardrailsPrompt augmentationMinimizing hallucinationsContext lengthDesign Applications Design Applications-14%14%Data Preparation Data Preparation-14%14%App Development App Development-30%30%

10、11EXAM SECTIONS 1EXAM SECTIONS 1-3 32024 Databricks Inc.All rights reservedDeploying via MLflowAccess controlModels in Unity CatalogVector search indexesModel serving endpointsGuardrails and maskingMalicious prompt protectionProblematic text in RAG data sourcesLegal and licensing concernsChoosing an

11、 LLMMonitoring and metricsPerformance evaluation with MLflowInference loggingLLM cost and controls for RAGAssembling and Deploying Assembling and Deploying Applications Applications-22%22%Governance Governance-8%8%Evaluation and Monitoring Evaluation and Monitoring-12%12%12EXAM SECTIONS 4EXAM SECTIO

12、NS 4-6 62024 Databricks Inc.All rights reserved13SAMPLE QUESTIONSAMPLE QUESTIONAfter changing the response-generating LLM in a RAG pipeline from GPT-4 to a model with a shorter context length that the company self-hosts,the Generative AI Engineer is getting the following error:error_code:BAD_REQUEST

13、,message:Bad request:rpc error:code=InvalidArgument desc=prompt token count(4595)cannot exceed 4096.What TWO solutions should the Generative AI Engineer implement without changing the response generating model?A.Use a smaller embedding model to generate embeddingsB.Reduce the maximum output tokens o

14、f the new modelC.Retrain the response generating model using ALiBiD.Decrease the chunk size of embedded documentsE.Reduce the number of records retrieved from the vector database2024 Databricks Inc.All rights reserved14EXAM PREPARATIONEXAM PREPARATION1.The Exam Guide,The Exam Guide,and The Exam Guid

15、e2.Databricks Courseware3.Review todays LLM-based APIs and tools4.Databricks Documentation2024 Databricks Inc.All rights reserved2024 Databricks Inc.All rights reserved15WHY GET WHY GET CERTIFIED ON CERTIFIED ON GENAI WITH GENAI WITH DATABRICKS?DATABRICKS?2024 Databricks Inc.All rights reserved16WHY CERTIFY?WHY CERTIFY?Confirm your skills in the most in-demand tech Show your organization you meet the Databricks standard now Get on Databricks GenAI path now for the future2024 Databricks Inc.All rights reserved2024 Databricks Inc.All rights reserved17Q&AQ&A