Study Guide · Bucket 6 of 7
AI & Machine Learning
2026 ka differentiator. AI ab optional skill nahi — expected hai. Yeh bucket 3 patterns pe based hai: Managed services, Serverless AI, aur Containers/GPU.
Context
3 patterns yaad rakho: (1) Managed services jaise Bedrock/SageMaker, (2) Serverless — Lambda + APIs event-driven AI workloads ke liye, (3) Containers — custom inference/training jab GPU chahiye ho.Amazon Bedrockmanaged foundation models
What
Foundation models (Claude, Llama, aur others) ko ek unified API ke through access — bina khud GPU infra manage kiye.
Why
Model train karne ki zaroorat nahi — bas API call karo, prompt bhejo, response lo. Yeh AI feature build karne ka sabse fast tareeka hai.
How
Bedrock API call karo (Boto3 ya AWS SDK se) -> model choose karo -> prompt bhejo -> response milta hai. Same pattern jaise koi bhi LLM API.
Skill
API calls karna — Python (Boto3) ya JavaScript (AWS SDK).
- Foundation models
- No training required
- Prompt -> API -> Response
Amazon SageMakercustom ML platform
What
End-to-end platform — build, train, tune, aur deploy custom ML models at scale. Data scientists yahan notebooks use karte hain.
Why
Jab pre-built model kaam na aaye aur khud ka custom model chahiye ho (specific business problem ke liye), SageMaker uska platform hai.
How
Notebook mein model develop karo -> SageMaker se train karo -> ek endpoint pe deploy karo -> application us endpoint ko API ki tarah call karta hai (Bedrock jaisa hi pattern — prompt in, response out).
Skill
High-level understanding kaafi hai — deep hands-on data-scientist roles ke liye zyada relevant hai.
- Notebooks
- Training jobs
- Endpoints
Amazon Q DeveloperAI coding assistant
What
VS Code, JetBrains, aur AWS Console mein built-in AI coding assistant — code likhne, AWS concepts samajhne, aur debug karne mein madad karta hai.
Why
Productivity multiplier hai — isko avoid nahi karna, use karna chahiye.
How
IDE mein extension install karo, inline suggestions aur chat se code likho/debug karo.
Skill
Tool use karna — install aur prompt karna.
- AI pair-programming
Textract, Rekognition, Comprehendpre-trained AI APIs
What
Textract = documents se text/table extract karta hai. Rekognition = image/video analysis. Comprehend = Natural Language Processing (sentiment, entities, etc.).
Why
Yeh task-specific pre-trained APIs hain — model train karne ki zaroorat nahi, seedha API call karo.
How
File (S3 se) API ko bhejo -> structured result milta hai (extracted text, detected objects, sentiment score, etc.).
Skill
API integration.
- Task-specific pre-trained models
Key Tasks — is bucket ko 'done' maanne se pehle
- Bedrock API call karo (Boto3) — ek prompt bhejo, response print karo
- SageMaker notebook explore karo — ek sample model dekho (deep training zaroori nahi)
- Amazon Q Developer VS Code mein install karo, use karke ek function likho
- Textract se ek document (PDF/image) se text extract karo