2026.04.10 / 동준상.넥스트플랫폼
(AWS SAA, AWS AIF, GCP GenAI Leader)
이번 포스트는 AI 네이티브 클라우드 실무활용을 위한 핸즈온 참고자료입니다.
사전 공통 준비사항
0. 크롬 브라우저 (구글 계정 동기화 필요) + Google Cloud 회원 가입 (신용카드 정보입력, 결제 아님, 프리티어 사용)
https://cloud.google.com/free
1. Google Skills 접속 (구글 클라우드 서비스 프리티어 핸즈온 랩)
https://www.skills.google/
2. 카탈로그 주제 검색 창에 ‘Vertex AI’입력
>> 무료 리소스 접근 가능한 Labs 선택
핸즈온 1단계: AI 네이티브 클라우드 시작하기
1.1 Vertex AI Studio 시작하기
(Get Started with Vertex AI Studio)
- Setup and requirements
- Task 1. Create applications from prompts
- Task 2. Design effective prompts
- Task 3. Engineer and manage prompts
- Task 4. Use multimodal prompts with Gemini
- Task 5. Generate media in Vertex AI Studio
1.2 Vertex AI 프롬프트 설계
(Prompt Design in Vertex AI: Challenge Lab)
Setup and requirements
Challenge Scenario
Task 1. Build a Gemini image analysis tool
Task 2. Build a Gemini tagline generator
Task 3. Experiment with image analysis code
Task 4. Experiment with tagline generation code
핸즈온 2단계: AI 네이티브 클라우드 활용하기
2.1 Vertex AI Vector Search 하이브리드 검색
(Create Hybrid Search With Vertex AI Vector Search)
Setup and requirements
Task 1. Open the notebook in Vertex AI Workbench and choose kernel
Task 2. Install packages, and configure the notebook.
Task 3. Prepare the dataset
Task 4. Create sparse embedding
Task 5. Create an index endpoint
Task 6. Create the hybrid index and deploy it to the Endpoint
Task 7. Run a hybrid query
2.2 Vertex AI Gemini API 기반 멀티모달 RAG 구현
(Multimodal Retrieval Augmented Generation (RAG) using the Gemini API in Vertex AI)
Setup and requirements
Task 1. Open the notebook in Vertex AI Workbench
Task 2. Set up the notebook
Task 3. Download custom Python utilities & required files
Task 4. Build metadata of documents containing text and images
Task 5. Text Search
Task 6. Image Search
Task 7. Multimodal retrieval augmented generation (RAG)
2.3 Vertex AI Gemini를 이용한 정보 신뢰성 검증
(Intro to Grounding with Gemini in Vertex AI)
Setup and requirements
Task 1. Open the notebook in Vertex AI Workbench
Task 2. Set up the notebook
Task 3. Grounding with Google Search results
Task 4. Create a Vertex AI Datastore
Task 5. Create a Vertex AI Search Application
Task 6. Grounding with custom documents and data
Task 7. Grounded chat responses
핸즈온 3단계: AI 네이티브 클라우드 응용하기
3.1 Vertex AI Gemini API를 이용한 세이프가드 구현
(Safeguarding with Vertex AI Gemini API)
Task 0. Setup and requirements
Task 1. Open Vertex AI Workbench instance
Task 2. Clone a course repo within your Vertex AI Workbench instance
Task 3. Safeguard with Gemini API
3.2 Vertex AI를 이용한 ADK 성능 평가
(Evaluate ADK agent performance using the Vertex AI Generative AI Evaluation Service)
Setup and requirements
Task 1. Prepare the environment in Vertex AI Workbench
Task 2. Read and run the rest of the notebook
핸즈온 랩 진행 예시 및 도움말
1.1 Vertex AI Studio 시작하기
Overview
Vertex AI Studio lets you quickly test and customize generative AI models so you can leverage their capabilities in your applications.
Vertex AI Studio를 사용하면 생성형 AI 모델을 신속하게 테스트하고 맞춤 설정하여, 모델의 기능을 애플리케이션에 활용할 수 있습니다.
Objectives
Create applications from prompts.
Design effective prompts.
Engineer and manage prompts.
Use multimodal prompts.
프롬프트를 활용한 애플리케이션 생성
효과적인 프롬프트 설계
프롬프트 엔지니어링 및 관리
멀티모달 프롬프트 사용
Setup and requirements
Open Link in Incognito Window
시크릿창에서 링크 열기
Task 1. Create applications from prompts
1.4 System instructions
You are an expert AI assistant for an insurance underwriting department.
Your primary goal is to help underwriters by accurately and concisely summarizing client information and highlighting potential risk factors.
Maintain a professional and objective tone.
Focus only on the information provided in the prompt. Do not invent details.
귀하는 보험 인수(Underwriting) 부서의 전문 AI 어시스턴트입니다.
귀하의 주요 목표는 고객 정보를 정확하고 간결하게 요약하고 잠재적 위험 요소를 강조함으로써 언더라이터(인수 심사역)를 돕는 것입니다.
전문적이고 객관적인 어조를 유지하십시오.
프롬프트에서 제공된 정보에만 집중하십시오. 세부 사항을 임의로 지어내지 마십시오.
1.5 main prompt
Customer Notes for ‘SafeHarbor Warehousing’:
“The applicant is seeking coverage for their 50,000 sq ft warehouse. The business is 5 years old. The building is a concrete tilt-up structure, originally built in 2010. They store a variety of non-hazardous dry goods.
Fire safety measures include a full sprinkler system, a centrally monitored fire alarm, and documented annual inspections by a certified third party.
Security measures include a 24/7 centrally monitored burglar alarm, comprehensive security camera coverage of the interior and exterior, a fully fenced perimeter, and nightly patrols by a contracted security guard service.
The company reports no major property or liability losses in their 5-year history. They have specifically asked to ensure their new automated shelving and retrieval system, installed last month, is adequately covered under the policy.”
Your Task:
- Briefly summarize the key details of the ‘SafeHarbor Warehousing’ business and its existing safety measures.
- Based only on the notes provided, identify any immediate questions an underwriter should ask or potential risk factors they should consider further.
Present the summary first, then the questions/risk factors as bullet points.
‘SafeHarbor Warehousing’ 고객 참고 사항:
“신청자는 50,000평방피트 규모의 창고에 대한 보험 보장을 원합니다. 사업체 업력은 5년입니다. 건물은 2010년에 완공된 콘크리트 틸트업(tilt-up) 구조입니다. 다양한 비위험 건조 화물을 보관합니다.
화재 안전 조치로는 전체 스프링클러 시스템, 중앙 모니터링 화재 경보기가 포함되며, 인증된 제3자 기관을 통한 연례 점검 기록을 보유하고 있습니다.
보안 조치로는 24시간 연중무휴 중앙 모니터링 방범 알람, 내부 및 외부 전체를 아우르는 보안 카메라, 울타리가 설치된 구역, 그리고 계약된 보안 업체의 야간 순찰 서비스가 포함됩니다.
해당 기업은 5년의 운영 기간 동안 주요 재산상 손실이나 배상 책임 손실이 없었다고 보고했습니다. 이들은 특히 지난달에 설치된 신규 자동 선반 및 회수 시스템(automated shelving and retrieval system)이 보험 정책에 따라 적절하게 보장되는지 확인해 줄 것을 요청했습니다.”
귀하의 작업:
‘SafeHarbor Warehousing’의 사업 내용과 기존 안전 조치의 핵심 세부 사항을 간략하게 요약하십시오.
제공된 참고 사항에만 근거하여, 언더라이터가 즉시 질문해야 할 사항이나 추가로 고려해야 할 잠재적 위험 요소를 식별하십시오.
요약을 먼저 제시한 후, 질문 및 위험 요소를 글머리 기호로 나열하십시오.
1.17 Chatbot section
New Customer Inquiry:
“Applicant ‘Coastal Goods Delivery’ has a fleet of 10 delivery vans, all equipped with GPS and telematics. They operate within a 100-mile radius of their depot. Drivers undergo annual safety training. They had one minor fender bender last year, no injuries, $1500 damages. What are the primary risk considerations?”
Please summarize key points and identify potential risks.
신규 고객 문의:
“신청자 ‘Coastal Goods Delivery’는 10대의 배송 밴을 보유하고 있으며, 모든 차량에는 GPS와 텔레매틱스 장비가 장착되어 있습니다. 이들은 기지로부터 100마일 반경 내에서 영업합니다. 운전자들은 연례 안전 교육을 받습니다. 작년에 부상자가 없는 경미한 접촉 사고가 한 건 있었으며, 피해액은 1,500달러였습니다. 주요 위험 고려 사항은 무엇입니까?”
핵심 사항을 요약하고 잠재적 위험을 식별해 주십시오.
Task 2. Design effective prompts
Zero-shot prompting
2.5 System instructions
You are an AI assistant specializing in parsing and extracting specific data points from unstructured insurance claim notifications.
Your goal is to identify and list key information accurately.
If a piece of information is not found, clearly state “Not found”.
Output the extracted information in a key: value format, with each key on a new line.
2.6 main prompt
Claim Notification Received:
“Hi team, just got a call from Mrs. Eleanor Vance, policy #POL458892. She reported a kitchen fire that occurred on May 12th, 2025, around 3 PM. The main damage seems to be to the oven and surrounding cabinets. She mentioned smoke damage in the kitchen and dining area too. She thinks the total damage might be around $7,500. Her contact is 555-0123. No injuries reported, thankfully.”
Extract the following:
- Policy Number
- Claimant Name
- Date of Loss
- Time of Loss
- Type of Loss
- Brief Description of Damage
- Estimated Loss Amount
- Injuries Reported
Few-Shot prompting
2.4 Examples interface: INPUT
Claim Notification Received:
“Email from John Sterling (policy POL77521) re: water damage at his shop. Happened sometime last night, May 10th, 2025. A pipe burst in the ceiling. Stockroom is flooded, some damage to inventory. He’s not sure on the cost yet, maybe $5k-$10k? No one was there, so no injuries.”
Extract the following:
- Policy Number
- Claimant Name
- Date of Loss
- Time of Loss
- Type of Loss
- Brief Description of Damage
- Estimated Loss Amount
- Injuries Reported
2.4 Examples interface: OUTPUT
Policy Number: POL77521
Claimant Name: John Sterling
Date of Loss: May 10th, 2025
Time of Loss: Night
Type of Loss: Water damage
Brief Description of Damage: Pipe burst in ceiling, stockroom flooded, some damage to inventory.
Estimated Loss Amount: $5,000 – $10,000
Injuries Reported: No
2.5 Re-add System Instructions
You are an AI assistant specializing in parsing and extracting specific data points from unstructured insurance claim notifications.
Your goal is to identify and list key information accurately.
If a piece of information is not found, clearly state “Not found”.
Output the extracted information in a key: value format, with each key on a new line.
2.6 {Input} Write value here
Claim Notification Received:
“Hi team, just got a call from Mrs. Eleanor Vance, policy #POL458892. She reported a kitchen fire that occurred on May 12th, 2025, around 3 PM. The main damage seems to be to the oven and surrounding cabinets. She mentioned smoke damage in the kitchen and dining area too. She thinks the total damage might be around $7,500. Her contact is 555-0123. No injuries reported, thankfully.”
2.7 Write a prompt
Extract the following data points from the provided claim notification:
- Policy Number
- Claimant Name
- Date of Loss
- Time of Loss
- Type of Loss
- Brief Description of Damage
- Estimated Loss Amount
- Injuries Reported
Experimenting with prompt configurations
2.3 Insurance Story
Write the first paragraph of a short story about a homeowner who just used a futuristic AI insurance app to file a claim. The claim was for a bizarre and unexpected incident.
Task 3. Engineer and manage prompts
3.3 System instructions
You are an insurance risk analyst assistant. Your task is to identify potential risk factors from a given scenario. Be concise.
3.3 main prompt
Scenario:
“The applicant, ‘The Fiery Grill,’ is a new upscale restaurant specializing in wood-fired oven pizzas and open-flame grilling. They have installed a brand new, custom-built fire suppression system for their cooking area, but it has not yet been certified by a third party. The restaurant plans to feature live acoustic music on weekend evenings and has a small, raised stage area. They also want to offer valet parking.”
Based on this scenario, list three primary risk factors an underwriter should consider.
Task 4. Use multimodal prompts with Gemini
4.7
- Provide a concise title for this image (under 5 words).
- Describe the image in one or two sentences.
- Extract all visible text from the image. Present the flight schedule as a clearly formatted list with columns for “Time” and “City”.
4.9
Based on the flight schedule shown in the image, what percentage of the listed flights depart before 11:30 AM? Show your calculation if possible.
Task 5. Generate media in Vertex AI Studio
Imagen – prompt area
A close-up, photorealistic image of a single honeybee collecting pollen from a vibrant purple lavender flower, with a softly blurred garden background.
Chirp – prompt area
Welcome to the world of generative AI on Google Cloud
