Building a Real-Time ML System
Together

Learn to build and deploy end-2-end ML systems using Python, Rust, LLMs and Kubernetes.

Join the Best Interactive ML Course On the Internet & Stand Out From the Crowd

Next Cohort: December 2 - December 26, 2024

✔️ Access every future cohort forever. No need to pay again.
✔️ 60+ hours of recorded sessions from the previous 2 cohorts
✔️ Full source code of a crypto price predictor system.

Wondering how real-time ML systems are built?

You’ve come to the right place!In this course, we will build, deploy, and scale a real-time ML predictor for algorithmic trading.

Who is this course for?

Are you an ML engineer, data scientist or a software engineer following hot trends in the industry but missing the hands-on training that would allow you to become a part of the change?I can help you bridge the gap!Build a REAL-TIME Prediction System for Trading with me in a live and interactive 3-weeks course that will change your career forever.It is time to stop following and start BUILDING.To succeed, you should:

  • Have experience writing code. We will use Python but if you know a different programming language, you can follow easily.

  • You are familiar with the Machine Learning fundamentals and you’ve trained at least one ML model before.

  • You are not afraid of hard work. (That will have a huge pay off, I promise.)

Why should you join?

Because when you join you get life-time access to:

  • Full source code of many real-time ML systems: so far a crypto price predictor, and a taxi trip duration à la Uber.

  • 36 hours of live, interactive coding sessions.

  • We meet 3 times a week, and we code for 3 hours. We do this for 4 weeks, hence 36 hours.

  • Direct communication with your instructor (that’s me, by the way).

  • A community of similar-minded ML builders eager to learn new things and solve real-world problems.

  • A list of project ideas you will be able to implement after finishing the course & feedback on your final project.

  • Lifetime access to all future cohorts' sessions. You can participate in as many iterations as you want.

  • The possibility to work with top companies. If you do a good job I will be more than happy to give visibility to your work on my social media accounts and recommend you to my clients.

The best part? You pay ONCE and get access to all live and recorded future sessions. That means you don’t have to stress if you skip anything in this cohort, you can join the future ones whenever you want :-)Forget about theory and passive learning, it is time to roll up your sleeves and become a part of an exciting and challenging project that will turn your abstract theoretical knowledge to ready-to-apply PRACTICE.

What will you learn?

This is what makes this program unique:

  • You’ll learn to design, build, deploy and scale microservice architectures that use Real-Time Machine Learning to continuously generate predictions.

  • You will learn to design any ML system using the universal Feature-Training-Inference Pipeline design.

  • You will learn how to build real-time ML services in Python, using best-of-breed Serverless tools, including a Kafka streaming data platform, a Feature Store, an Experiment Tracker, a Model Registry, and a Compute Platform à la Kubernetes.

Check the program syllabus →

Ready for a serious knowledge upgrade?

Join the program TODAY and get a life-time access to building the hottest real-time ML project out there!
Early bird pricing until 1st of June

Still have doubts? Reach out on social media and I'll be happy to help

$600

$420

🚨 30% discount ending soon

You pay once and get lifetime access to all future sessions. No restrictions.

  • Build a real-time ML system from scratch over 4 weeks.

  • Access to the full source code of an ML predictor for algorithmic trading.

  • Learn to integrate Large Language Models in an end-2-end ML system.

  • 36+ hours of live, interactive coding sessions.

  • Office hours sessions to solve any doubts you may have.

  • Direct communication with your instructor (that’s me, by the way).

  • A community of similar-minded ML builders eager to learn new things and solve real-world problems.

  • A list of project ideas you will be able to implement after finishing the course & feedback on your final project.

  • Lifetime access to all future cohorts' sessions. You can participate in as many iterations as you want.

  • The possibility to work with top companies. If you do a good job I will be more than happy to give visibility to your work on my social media accounts and recommend you to my clients.

Current cohort

All coding sessions and office hours start at 11:00 AM Central European Time.All coding sessions and office hours are recorded and uploaded If you cannot make it to an office hour, but would like to ask something, please ask on the Discord Community, and I will answer during the session.

  • Monday → Live session, 3 hours

  • Tuesday → Individual work

  • Wednesday → Office hours + coding (3 hours)

  • Thursday → Live session, 3 hours

  • Friday → Individual work

Program syllabus

Session 1 - How to build a real-time ML system

  • Problem framing

  • System architecture

  • The Feature-Training-Inference pipeline design

  • How to iteratively build modular ML systems

  • Open-source development tools and Serverless Infrastructure

  • Feature pipeline components

  • Docker fundamentals

  • Let's develop, containerize and run our first microservice using Python and Docker.

  • Makefile to streamline the development workflow

  • Automatic code linting and formatting

Session 2 - The Feature pipeline

  • Real-time data ingestion from Websocket APIs

  • Stateful real-time data transformations in Python

  • Reading and writing data to Kafka in Python

  • How to write modular Python code, that is easy to debug and test.

  • What is a Feature Store and why do we need one?

  • Backfilling of historical data

  • What is offline-online skew and how to avoid it

  • Kafka topics, partitions and consumer groups

  • Horizontal scaling of data transformations with Docker and Kafka.

Session 3 - The Training pipeline

  • Iterative development of ML models.

  • Training and evaluation of time-series forecasts

  • Model-dependent transformations

  • Feature engineering using trading technical indicators

  • Target metric parametrisation

  • Integration of experiment tracking

  • What is a model registry and how to integrate it to the training pipeline

  • Dockerization for reproducible runs

Session 4 - The Inference pipeline

  • Real-time ML model prediction in Python

  • How to build a streaming microservice for live predictions

  • Visualization of prediction in Streamlit

  • Connecting the inference pipeline to the model registry

Session 5 - Deployment and monitoring

  • Deployment of microservices à la Kubernetes

  • Feature latency monitoring using Prometheus and Grafana

  • Feature quality validation

  • Model error monitoring

  • Model re-training

Session 6 - System improvements

This session is open-ended at the moment. Based on your feedback from the previous session I will decide what we cover here.

  • Feature engineering hyperparameter optimization

  • Horizontal scaling of feature engineering services

  • Real-time visualization of features with Streamlit

  • Open-ended

Hey! I'm Pau.

I am a Mathematician turned Machine Learning Engineer, turned Machine Learning educator.I love solving problems, so much so that I participated in the International Mathematical Olympiad. And ML is all about solving problems.I started my career as a Quantitative Analyst for Erste Bank more than 10 years ago and then worked as a Data Scientist in a mobile gaming company Nordeus (now Take-Two Interactive).There, I had my first real-world ML project and decided this is something I want to do.Since then, I’ve worked as a freelancer at Toptal on projects involving self-driving cars, finance, delivery aps, time-series prediction for online retail, health insurance providers, and more.Two years ago, I started sharing what I know with my community on Twitter/X and LinkedIn. My first course The Real-World Machine Learning Tutorial has more than 500 happy students and I keep sharing free content on my social media.Over the years, I’ve realized that the only way to learn is to actually DO it and in this course I will be sharing everything I know about real-time ML projects with you.This course gives you practical, ready-to-use knowledge that will change your career for the better.I can’t wait to meet you. See you at the course!

© realworldml.net. All rights reserved.