The focus of this project is on deploying a machine learning model to AWS. Although I did make predictions, the main objective is to set up all the necessary AWS resources (S3, SageMaker, Lambda, API Gateway) to host my model in a scalable, serverless environment.

• Amazon S3 – for storing the training data (train.csv).
• Amazon SageMaker – for training my XGBoost model on the bank dataset and hosting a real-time inference endpoint.
• AWS Lambda – for providing inference logic (calls the SageMaker endpoint).
• Amazon API Gateway – for creating an HTTP endpoint that triggers the Lambda function.

1. Clone or download this project.
2. Open a terminal in the project folder (where train_and_deploy.py is located).
3. Install necessary libraries (e.g., boto3, sagemaker, pandas, etc.) if not already available.
4. Run the script to train and deploy:

   python3 train_and_deploy.py

   • This script will:
     • Create an S3 bucket (if it does not exist).
     • Upload train.csv to S3.
     • Launch an XGBoost training job in SageMaker.
     • Deploy the resulting model to a SageMaker endpoint.
     • Print out the SageMaker endpoint name at the end (e.g., xgboost-2025-01-26-16-02-04-923).

1. Go to the AWS Lambda Console and create a new function (e.g., the default one is given as lambda_function.py).

2. Paste the following code into the Lambda console. Make sure to replace the ENDPOINT_NAME with the actual endpoint name printed by train_and_deploy.py if it differs.

   import json
   import boto3
   
   # Replace with your actual SageMaker endpoint name
   ENDPOINT_NAME = "xgboost-2025-01-26-16-02-04-923"
   
   def lambda_handler(event, context):
       # 1. Parse all input features from queryStringParameters
       query_params = event.get('queryStringParameters', {})
       
       # List of all required feature names in the correct order
       required_features = [
           "age", "campaign", "pdays", "previous", "no_previous_contact", "not_working",
           "job_admin.", "job_blue-collar", "job_entrepreneur", "job_housemaid",
           "job_management", "job_retired", "job_self-employed", "job_services",
           "job_student", "job_technician", "job_unemployed", "job_unknown",
           "marital_divorced", "marital_married", "marital_single", "marital_unknown",
           "education_basic.4y", "education_basic.6y", "education_basic.9y", "education_high.school",
           "education_illiterate", "education_professional.course", "education_university.degree",
           "education_unknown", "default_no", "default_unknown", "default_yes",
           "housing_no", "housing_unknown", "housing_yes", "loan_no", "loan_unknown",
           "loan_yes", "contact_cellular", "contact_telephone", "month_apr", "month_aug",
           "month_dec", "month_jul", "month_jun", "month_mar", "month_may", "month_nov",
           "month_oct", "month_sep", "day_of_week_fri", "day_of_week_mon", "day_of_week_thu",
           "day_of_week_tue", "day_of_week_wed", "poutcome_failure", "poutcome_nonexistent",
           "poutcome_success"
       ]
   
       # Extract values for required features and ensure all are present
       data = [float(query_params.get(feature, 0)) for feature in required_features]
   
       # 2. Format as CSV (single row)
       payload = ",".join(map(str, data))
   
       # 3. Invoke SageMaker endpoint
       runtime = boto3.client('sagemaker-runtime')
       try:
           response = runtime.invoke_endpoint(
               EndpointName=ENDPOINT_NAME,
               ContentType='text/csv',
               Body=payload
           )
   
           # 4. Parse prediction output
           result = response['Body'].read().decode('utf-8')
           prediction = float(result.strip())
   
           return {
               "statusCode": 200,
               "body": json.dumps({
                   "prediction": prediction
               }),
               "headers": {
                   "Content-Type": "application/json"
               }
           }
   
       except Exception as e:
           # Handle errors gracefully
           return {
               "statusCode": 500,
               "body": json.dumps({"error": str(e)})
           }

3. Deploy the Lambda function changes (click "Deploy" in the Lambda console).

1. Create an HTTP API in the Amazon API Gateway console.
2. Add a Lambda Integration:
   • Choose "Lambda" as the integration type.
   • Select the Lambda function (BankInferenceLambda).
3. Define a Route:
   • For example, GET /BankInferenceLambda.
   • Attach the newly created Lambda integration to that route.
4. Obtain the Invoke URL shown in the API Gateway console.
   For example:

   https://ou38vl9wcj.execute-api.us-east-1.amazonaws.com/BankInferenceLambda
   

In a browser or using curl, append query parameters for each feature. For example:

https://ou38vl9wcj.execute-api.us-east-1.amazonaws.com/BankInferenceLambda
    ?age=56
    &campaign=1
    &pdays=999
    &previous=0
    &no_previous_contact=1
    &not_working=0
    &job_admin.=0
    &job_blue-collar=0
    &job_entrepreneur=1
    &job_housemaid=0
    &job_management=0
    &job_retired=0
    &job_self-employed=0
    &job_services=0
    &job_student=0
    &job_technician=0
    &job_unemployed=0
    &job_unknown=0
    &marital_divorced=1
    &marital_married=0
    &marital_single=0
    &marital_unknown=0
    &education_basic.4y=1
    &education_basic.6y=0
    &education_basic.9y=0
    &education_high.school=0
    &education_illiterate=0
    &education_professional.course=0
    &education_university.degree=0
    &education_unknown=1
    &default_no=0
    &default_unknown=0
    &default_yes=1
    &housing_no=0
    &housing_unknown=0
    &housing_yes=1
    &loan_no=0
    &loan_unknown=0
    &loan_yes=0
    &contact_cellular=0
    &contact_telephone=1
    &month_apr=0
    &month_aug=0
    &month_dec=0
    &month_jul=0
    &month_jun=0
    &month_mar=0
    &month_may=1
    &month_nov=0
    &month_oct=0
    &month_sep=0
    &day_of_week_fri=0
    &day_of_week_mon=0
    &day_of_week_thu=0
    &day_of_week_tue=0
    &day_of_week_wed=1
    &poutcome_failure=0
    &poutcome_nonexistent=0
    &poutcome_success=0

When you visit this URL, you should see a JSON response like:

{
  "prediction": 0.123456
}

(The numerical value depends on the model’s prediction.)

1. Trained the model by running python3 train_and_deploy.py, which deployed a SageMaker endpoint.
2. Set Up a Lambda Function to handle incoming requests and call the endpoint with CSV-formatted features.
3. Exposed that Lambda function via API Gateway to produce a public URL.
4. Tested the live endpoint by passing query parameters in a browser or via curl.

The project demonstrates how to deploy an ML model in a serverless AWS environment, emphasizing how these AWS services connect rather than focusing solely on the prediction results.