Why Serverless for ML?

123456789101112
import json
import os
import boto3
import numpy as np

# Initialize model OUTSIDE handler for container reuse
# This code runs once per cold start, not per invocation
MODEL = None
S3_CLIENT = boto3.client('s3')

def load_model():
    """Load model from S3 or Lambda Layer - runs once per container"""

123456789101112
import json
import os
import boto3
from functools import lru_cache
import numpy as np

# Global model cache - persists across warm invocations
_model = None
_model_version = None

def get_model(version: str):
    """Load model with caching and version checking."""

123456789101112
import json
import pickle
import os
from typing import Dict, Any

# Global model cache - persists across warm invocations
MODEL = None

def load_model():
    global MODEL
    if MODEL is None:
        model_path = os.path.join(os.path.dirname(__file__), 'model.pkl')

123456789101112
{
  "Comment": "Serverless ML Inference Pipeline with Error Handling",
  "StartAt": "ValidateInput",
  "States": {
    "ValidateInput": {
      "Type": "Task",
      "Resource": "arn:aws:lambda:us-east-1:123456789:function:validate-input",
      "Next": "FetchFeatures",
      "Catch": [{
        "ErrorEquals": ["ValidationError"],
        "Next": "HandleValidationError"
      }],

123456789101112
import * as cdk from 'aws-cdk-lib';
import * as lambda from 'aws-cdk-lib/aws-lambda';
import * as apigateway from 'aws-cdk-lib/aws-apigateway';
import * as dynamodb from 'aws-cdk-lib/aws-dynamodb';

export class ServerlessMlStack extends cdk.Stack {
  constructor(scope: cdk.App, id: string, props?: cdk.StackProps) {
    super(scope, id, props);

    // Feature cache for low-latency lookups
    const featureCache = new dynamodb.Table(this, 'FeatureCache', {
      partitionKey: { name: 'userId', type: dynamodb.AttributeType.STRING },