Complete AWS Lambda + SageMaker Tutorial: Serverless ML Inference

Combining AWS Lambda with SageMaker enables serverless, scalable, and cost-effective ML inference. This tutorial covers patterns for integrating Lambda functions with SageMaker endpoints for production ML applications.

Why Lambda + SageMaker?

• Serverless: No infrastructure management
• Cost-effective: Pay per invocation
• Scalable: Automatic scaling to demand
• Flexible: Multiple integration patterns
• Event-driven: Trigger inference from any AWS event

• Real-time API inference
• Batch processing triggers
• Event-driven ML pipelines
• Cost-optimized inference
• Multi-model serving

Prerequisites

pip install boto3 sagemaker

AWS CLI configured
aws configure

Pattern 1: Lambda Invoking SageMaker Endpoint

1. Deploy SageMaker Endpoint

import sagemaker
from sagemaker.sklearn import SKLearnModel

session = sagemaker.Session()
role = sagemaker.getexecutionrole()

Deploy model
model = SKLearnModel(
    modeldata="s3://bucket/model/model.tar.gz",

    role=role,
    frameworkversion="1.0-1",
    pyversion="py3"

)

predictor = model.deploy(
    initialinstancecount=1,

    instancetype="ml.m5.large",
    endpointname="sklearn-endpoint"

)

# lambdafunction.py
import json
import boto3
import os

Initialize SageMaker runtime client
sagemakerruntime = boto3.client("sagemaker-runtime")


ENDPOINTNAME = os.environ.get("SAGEMAKERENDPOINT", "sklearn-endpoint")


def lambdahandler(event, context):
    """Invoke SageMaker endpoint for inference."""
    try:
        # Parse input
        if "body" in event:
            body = json.loads(event["body"])
        else:
            body = event

        features = body.get("features", [])

        # Prepare payload
        payload = json.dumps({"features": features})

        # Invoke endpoint
        response = sagemakerruntime.invokeendpoint(
            EndpointName=ENDPOINTNAME,

            ContentType="application/json",
            Body=payload
        )

        # Parse response
        result = json.loads(response["Body"].read().decode())

        return {
            "statusCode": 200,
            "headers": {"Content-Type": "application/json"},
            "body": json.dumps({
                "prediction": result,
                "endpoint": ENDPOINTNAME
            })
        }

    except Exception as e:
        return {
            "statusCode": 500,
            "body": json.dumps({"error": str(e)})
        }

3. Deploy Lambda with SAM

# template.yaml
AWSTemplateFormatVersion: '2010-09-09'
Transform: AWS::Serverless-2016-10-31

Resources:
  InferenceFunction:
    Type: AWS::Serverless::Function
    Properties:
      Handler: lambdafunction.lambdahandler
      Runtime: python3.9
      Timeout: 30
      MemorySize: 256
      Environment:
        Variables:
          SAGEMAKERENDPOINT: sklearn-endpoint

      Policies:
Version: '2012-10-17'
          Statement:

Effect: Allow
              Action:

sagemaker:InvokeEndpoint
              Resource: '*'

      Events:
        Api:
          Type: Api
          Properties:
            Path: /predict
            Method: post

# Deploy
sam build
sam deploy --guided

# Dockerfile
FROM public.ecr.aws/lambda/python:3.9

Install dependencies
COPY requirements.txt .
RUN pip install -r requirements.txt