Data Science with Python and FastAPI is a course in predictive analytics with Python, from CRISP-DM and data preparation with pandas to building models with scikit-learn, tracking them with MLflow and serving them through a FastAPI prediction API backed by a database. It suits data scientists, Python developers and learners who have already completed a basic Python course.
In many organisations data science work stops at the notebook. A model that forecasts sales or risk is finished but never used, because other systems cannot call it, nobody knows which version is best, and predictions are not kept for later review. Data science delivers value when models are built properly and handed over as a service that other systems can call securely.
This course takes learners through predictive analytics in Python from raw data to an API that works in practice. It starts by planning the project with CRISP-DM, preparing and cleaning data with pandas 3 and building models with scikit-learn 1.9, including regression, clustering, decision trees, association rules and neural networks, inside reusable pipelines. Learners then save and track models with MLflow and build a RESTful API with FastAPI that loads the model to make predictions, connects to a database with SQLModel, is protected with JWT, tested with pytest and run with Docker. The course closes with an end-to-end prediction system. (5 days, 6 hours per day, 30 hours in total, Intermediate level.)
What you’ll gain
Plan a predictive analytics project following the CRISP-DM process
Prepare, clean and explore data correctly with pandas 3 and charts
Build pipelines and models with scikit-learn, including regression, clustering, decision trees and neural networks
Evaluate models properly with cross-validation and tune hyperparameters
Save, compare and version models with skops and MLflow
Build a RESTful API with FastAPI that serves predictions from a model
Connect a database with SQLModel for CRUD operations and a prediction history
Protect the API with JWT, test it with pytest and run the system with Docker
Who this course is for
Data scientists and data analysts who want to take models out of the notebook and into real use
Python developers who want to add machine learning and API skills
Data engineers and IT teams who look after their organisation's prediction systems
Learners who have completed basic Python and want to move on to full data science work
Prerequisites
Basic Python, or completion of an introductory Python course
Basic statistics such as mean, standard deviation and relationships between variables
Basic SQL and an understanding of HTTP and JSON will help a great deal
A laptop that can run Python, VS Code and Docker Desktop, with permission to install software
Curriculum
Course Details
This is the 2026 edition of the Data Science with Python and FastAPI course. It keeps the original core of predictive analytics in Python and a FastAPI RESTful API with database access, rewritten for pandas 3, scikit-learn 1.9 and the current FastAPI. It runs for 5 days, 6 hours per day (30 hours in total, 09:00-16:00), as lectures with labs that build on fictional customer and sales data throughout. Intermediate level. Compared with the previous
edition it adds pipelines, cross-validation, gradient boosting, MLflow, skops, lifespan events, SQLModel, JWT, pytest and Docker. The course focuses on the path from data to a prediction API, unlike Data Science with Python, which concentrates on modelling, and the other FastAPI courses, which cover general API development. All tools are open source and no GPU is needed. Learners take home a lab guide, notebooks and code for every lab, and the prediction API project with docker-compose files.
Day 1Data and Predictive Foundations
Section 1: Lab: Predictive Modelling and CRISP-DM
What predictive analytics is and which business problems suit prediction
The six stages of CRISP-DM, from understanding the problem to deployment
The path a model takes from a notebook to a service other systems can call
Install Python and manage the project with uv, with JupyterLab or VS Code
Lab: plan a customer churn prediction project with CRISP-DM
Section 2: Lab: Working with Data in pandas 3
Read data from CSV, Excel and SQL databases into a DataFrame
What is new in pandas 3: Copy-on-Write and the str data type by default
Select, filter and create columns with loc and assign
Join tables with merge and summarise with groupby and pivot_table
Lab: build a one-row-per-customer table from several data sources
Section 3: Lab: Data Cleaning and EDA
Find and handle missing values, duplicates and wrong data types
Detect outliers and decide whether to remove, fix or keep them
Handle dates, text and categories
Explore the data with Matplotlib and seaborn charts
Lab: clean the customer data and summarise the first findings
Section 4: Statistics for Predictive Modelling
Distributions, central tendency, spread and skewness
Correlation, and the difference between association and cause
Splitting data into train and test sets properly, including stratified splits
What data leakage is and where it tends to creep in
Lab: test simple hypotheses and set a baseline for the model
Day 2Regression, Classification and Evaluation
Section 5: Lab: Feature Engineering and Pipelines
Encode categories with OneHotEncoder and scale numeric values
Handle columns of different types together with ColumnTransformer
Keep preparation and model in a single pipeline to prevent data leakage
Create new features from dates, behaviour and running totals
Lab: a pipeline that takes raw data and makes it ready for the model
Section 6: Lab: Linear Regression
How linear regression works and how to interpret its coefficients
Evaluate with MAE, RMSE and R²
Reduce overfitting with Ridge and Lasso
Check the residuals to see where the model goes wrong
Lab: forecast a store's weekly sales
Section 7: Lab: Logistic Regression and Evaluation
Logistic regression for predicting classes and probabilities
Confusion matrix, precision, recall, F1 and ROC AUC
Choose a threshold that matches the business cost
Deal with imbalanced data using class_weight and suitable metrics
Lab: predict which customers are likely to cancel
Section 8: Cross-Validation and Hyperparameters
Why a single train and test split is not enough
K-fold and stratified cross-validation with cross_validate
Tune hyperparameters with GridSearchCV and RandomizedSearchCV
Compare several models fairly on the same data
Lab: find the best model and settings for the churn problem
Day 3Clustering, Trees and Neural Networks
Section 9: Lab: Data Clustering
Unsupervised learning and customer segmentation problems
K-Means and choosing the number of clusters with the elbow and silhouette methods
DBSCAN and HDBSCAN for clusters that are not round
Describe each cluster so the business can act on it
Lab: segment customers by purchasing behaviour with RFM
Section 10: Lab: Decision Trees and Ensembles
How a decision tree works and how to read rules from it
Limit tree depth to reduce overfitting
Random Forest and HistGradientBoosting for tabular data
Explain models with feature importance and permutation importance
Lab: compare a decision tree with gradient boosting on the same problem
Section 11: Lab: Association Rules Discovery
Market basket analysis and products bought together
Support, confidence and lift
Find rules with Apriori and FP-Growth from mlxtend
Filter for useful rules and turn them into business recommendations
Lab: find products to promote together from shopping basket data
Section 12: Lab: Introduction to Neural Networks
Neurons, layers, activation functions and learning through backpropagation
Build neural networks with MLPClassifier and MLPRegressor
Scale the data, use early stopping and read learning curves
When tabular data is better served by gradient boosting, and when to move on to deep learning
Lab: compare a neural network with the models built earlier
Day 4From Model to Prediction API
Section 13: Lab: Saving and Tracking Models with MLflow
Save models with skops, which is safer than pickle and joblib
Log parameters, metrics and model files for each experiment with MLflow Tracking
Compare experiments in the MLflow UI
Register models in the Model Registry and give the production version an alias
Lab: choose the best model and prepare it for the API
Section 14: Lab: FastAPI Fundamentals
Declare GET, POST, PUT and DELETE endpoints
Take values from the path, query string and request body with Pydantic v2
Test through Swagger UI at /docs and with Postman or Bruno
Split routes into several files with APIRouter and configure CORS
Lab: a customer data API with automatic OpenAPI documentation
Section 15: Lab: A Prediction API with FastAPI
Load the model once at start-up with a lifespan handler
Design Pydantic schemas that match the features the model expects
Single and batch prediction endpoints that return probabilities
Handle malformed input and return clear error messages
Lab: a churn prediction API that uses the model from the registry
Section 16: Lab: Connecting a Database with SQLModel
The ORM idea and connecting FastAPI to PostgreSQL through SQLModel
Create models, sessions and database dependencies
Write GET, POST, PUT and DELETE endpoints that read and write real data
Store a prediction history with the model version for later review
Lab: customer CRUD and a prediction log table
Day 5Security, Quality and the Complete System
Section 17: Lab: Authentication and Authorisation
The OAuth2 password flow and JSON Web Tokens with PyJWT
Hash passwords with pwdlib using Argon2
Role-based permissions such as viewer, user and model administrator
API keys for other systems that call the prediction service
Lab: protect the prediction endpoints and customer data
Section 18: Lab: Testing and API Quality
Write tests with pytest and FastAPI's TestClient
Test the model contract, such as input types and the range of outputs
Use a test database separate from the live system
Logging and measuring API response times
Lab: a test suite that runs before every new release
Section 19: Lab: Docker and Looking After Models in Use
Build a Docker image of the API and combine it with PostgreSQL in Docker Compose
Configure the system with environment variables and pydantic-settings
Monitor whether incoming data drifts away from the training data
Plan retraining and switch model versions through an alias without downtime
Lab: run the whole prediction system with a single command
Section 20: Workshop: Capstone End-to-End Prediction System
Choose a brief such as sales forecasting, customer risk or customer segmentation
Go all the way from data preparation to building the model and logging it in MLflow
Hand it over as an API with a database, authentication and a test suite
Present the work, review it together and wrap up with a go-live checklist
Schedule & training options
For individuals — public rounds
No public rounds are open right now. Join the waiting list and we will contact you first when the next round opens, or ask us on LINE. Or call 02-570-8449 or 088-807-9770
CRISP-DM is a standard framework for data mining and analytics projects with six stages: business understanding, data understanding, data preparation, modelling, evaluation and deployment. The stages are iterative, so weak evaluation results can send you back to prepare the data again. In this course the deployment stage is done for real by serving the model as an API through FastAPI.
What is the difference between linear regression and logistic regression?
Linear regression predicts a continuous number, such as sales or property values, by fitting the line that best describes the relationship between variables, and is evaluated with MAE or RMSE. Logistic regression predicts a category or the likelihood of an event, such as whether a customer will cancel, giving a probability between 0 and 1 and then applying a threshold to decide the class. It is evaluated with precision, recall or ROC AUC.
How do you make a predictive model available to other systems through FastAPI?
Keep the data preparation and the model in a single scikit-learn pipeline, then save it with skops or register it in the MLflow Model Registry. In the FastAPI app, load the model once at start-up with a lifespan handler, define a Pydantic schema for the input data, and write a POST endpoint that calls predict and returns the result as JSON with probabilities. Test it through the Swagger page at /docs and write a test suite with pytest.
Who is the Data Science with Python and FastAPI course for, and what should I know beforehand?
It suits data scientists and data analysts who want to take models out of the notebook and into real use, Python developers who want to add machine learning skills, and data engineers or IT teams who look after prediction systems. You should know basic Python or have completed an introductory Python course and understand basic statistics. Some SQL and an understanding of HTTP and JSON will help a great deal.
What will I be able to do after the Data Science with Python and FastAPI course?
You will be able to plan projects with CRISP-DM, prepare and clean data with pandas 3, build models with scikit-learn including linear and logistic regression, clustering, decision trees, gradient boosting, association rules and neural networks, evaluate them with cross-validation, track and version them with MLflow, and build a FastAPI prediction API connected to a database with SQLModel, protected with JWT, tested with pytest and run with Docker. Participants who complete the training receive a Certificate of Completion with its own number, which can be verified online.
Can the Data Science with Python and FastAPI course be run in-house for our team?
Yes. It can be run in-house for your team, with examples adapted to your own work, such as building models and a prediction API for sales or risk problems close to your real cases. You can request a quotation from this page, or contact us on LINE @itgenius or by phone on 02-570-8449.
Python is one of the most widely used programming languages today, in web development, automation, data work, IoT and AI. Its readable syntax and gentle learning curve make it the first language of choice for many schools and organisations. Yet many beginners…
Many people can write basic Python but stay stuck with short scripts. When they need programs that run several tasks at once, talk over a network, send email, store data in a database or share code with a team, they are no longer sure how to write them.…
Machine learning is still at the heart of AI work in organisations, from sales forecasting and customer screening to document categorisation and anomaly detection, and a great deal of tabular data work is still handled better and more cheaply by classic…
18 hours3 days
7,650THB8,500 THB
View details
Data Science with Python and FastAPI30 hrs · 5 days
