Why REST API serving with FastAPI in MLOps? - Purpose & Use Cases

The Big Idea

What if your model could answer questions instantly without you lifting a finger?

The Scenario

Imagine you have a machine learning model that you want to share with your team or users. You try to send predictions by manually running scripts and emailing results every time someone asks.

The Problem

This manual way is slow and frustrating. You must run scripts each time, handle different requests by hand, and it's easy to make mistakes or miss requests. It's like answering the phone for every question instead of having a helpful assistant.

The Solution

FastAPI lets you create a REST API quickly and easily. It acts like a smart assistant that listens for requests, runs your model automatically, and sends back answers instantly. No more manual work or delays.

Before vs After

✗ Before

Run script.py
Check output
Email results

✓ After

from fastapi import FastAPI
from pydantic import BaseModel

app = FastAPI()

class InputData(BaseModel):
    # define your input fields here
    pass

@app.post('/predict')
async def predict(data: InputData):
    return model.predict(data)

What It Enables

With FastAPI, your model becomes instantly accessible to anyone, anytime, through simple web requests.

Real Life Example

A data scientist builds a model and uses FastAPI to serve predictions so a web app can show real-time recommendations to users without delays.

Key Takeaways

Manual sharing of model results is slow and error-prone.

FastAPI automates serving your model as a web service.

This makes your model easy to use and scales effortlessly.

Practice

1. What is the main purpose of using FastAPI in serving machine learning models?

easy

A. To train machine learning models faster

B. To create fast and simple REST APIs for model serving

C. To store large datasets efficiently

D. To visualize model performance graphs

Solution

1. Step 1: Understand FastAPI's role

   FastAPI is a web framework used to build APIs quickly and simply.

2. Step 2: Connect to model serving

   It is commonly used to serve machine learning models via REST APIs for easy access.

3. Final Answer:

   To create fast and simple REST APIs for model serving -> Option B

4. Quick Check:

   FastAPI = REST API serving [OK]

Hint: FastAPI is for building APIs, not training or storage [OK]

Common Mistakes:

• Confusing API serving with model training
• Thinking FastAPI handles data storage
• Assuming FastAPI is for visualization

2. Which of the following is the correct way to define a GET endpoint in FastAPI?

easy

A. @app.route('/items', method='GET')
def read_items():
return {'items': []}

B. @app.post('/items')
def read_items():
return {'items': []}

C. @app.get('/items')
def read_items():
print('items')

D. @app.get('/items')
def read_items():
return {'items': []}

Solution

1. Step 1: Identify correct decorator for GET

   FastAPI uses @app.get() to define GET endpoints.

2. Step 2: Check function returns JSON response

   The function should return a dictionary to send JSON; print statement does not return data.

3. Final Answer:

   @app.get('/items')\ndef read_items():\n return {'items': []} -> Option D

4. Quick Check:

   @app.get() + return dict = correct GET endpoint [OK]

Hint: Use @app.get() and return dict for GET endpoints [OK]

Common Mistakes:

• Using @app.post() for GET endpoints
• Using print instead of return
• Using Flask-style @app.route() syntax

3. What will be the output when you call the following FastAPI endpoint?

@app.get('/hello')
def say_hello():
    return {'message': 'Hello, FastAPI!'}

medium

A. {\"message\": \"Hello, FastAPI!\"}

B. Hello, FastAPI!

C. Error: Missing return type

D. 404 Not Found

Solution

1. Step 1: Analyze endpoint return value

   The function returns a dictionary with key 'message' and value 'Hello, FastAPI!'.

2. Step 2: Understand FastAPI response format

   FastAPI automatically converts dict to JSON response with the same structure.

3. Final Answer:

   {"message": "Hello, FastAPI!"} -> Option A

4. Quick Check:

   Return dict = JSON response with same keys [OK]

Hint: Return dict from endpoint gives JSON response [OK]

Common Mistakes:

• Expecting plain string instead of JSON
• Thinking missing return type causes error
• Assuming endpoint path is incorrect

4. Identify the error in this FastAPI POST endpoint code:

@app.post('/predict')
def predict(data: dict):
    return {'result': data['value'] * 2}

medium

A. Missing request body declaration with Pydantic model

B. Incorrect HTTP method, should be GET

C. Function missing return statement

D. Syntax error in decorator

Solution

1. Step 1: Check parameter type for POST data

   FastAPI requires request body to be declared with Pydantic models or Body for parsing JSON.

2. Step 2: Understand why dict alone is insufficient

   Using plain dict as parameter does not parse JSON body automatically, causing validation error.

3. Final Answer:

   Missing request body declaration with Pydantic model -> Option A

4. Quick Check:

   POST body needs Pydantic model or Body [OK]

Hint: Use Pydantic models for POST request bodies [OK]

Common Mistakes:

• Using plain dict instead of Pydantic model
• Confusing POST with GET method
• Forgetting to return a response

5. You want to serve a machine learning model prediction via FastAPI. Which approach correctly handles input validation and prediction?

from pydantic import BaseModel

class InputData(BaseModel):
    feature1: float
    feature2: float

@app.post('/predict')
def predict(data: InputData):
    result = model.predict([[data.feature1, data.feature2]])
    return {'prediction': result[0]}

What is the main advantage of this design?

hard

A. It skips input validation for faster response

B. It trains the model on each request

C. It validates input data types automatically before prediction

D. It returns raw model object instead of prediction

Solution

1. Step 1: Understand Pydantic model role

   InputData class validates that feature1 and feature2 are floats before function runs.

2. Step 2: Connect validation to prediction safety

   This prevents invalid data from reaching model.predict, avoiding runtime errors.

3. Final Answer:

   It validates input data types automatically before prediction -> Option C

4. Quick Check:

   Pydantic = automatic input validation [OK]

Hint: Use Pydantic models to validate inputs before prediction [OK]

Common Mistakes:

• Thinking model retrains on each request
• Skipping validation causes errors
• Returning model object instead of prediction