DS413
Building ML solutions: From System Design to Deployment
Faculty Profiles
Alexander Guschin
Industrial Head of Machine Learning at Central University
Olga Filippova
Lead Data Scientist at X5 Group, Russia
Course length
Duration
Total hours
Credits
Language
Course type
Fee for single course
Fee for degree students
Skills you’ll learn
Overview
This class will help you understand how complex ML systems are built and will be part of your portfolio that you can showcase and reason about. In three weeks, we're going to learn the principles of ML System Design and apply them to build an ML solution for a real-life problem. The problem will be detecting bots that pretend to be human on https://youare.bot/ platform. The problem is widely spread and is being solved in many social platforms and other places where communications between humans happen. We’ll deploy the bot classifier service and connect it to the real production system on https://youare.bot to get feedback, debug and improve our ML solution. To learn the problem from the other side of the fence, we’ll deploy our own bots designed to trick humans and classifiers. Eventually, we’ll have a hackathon to see who can design the best classifier and the best bot.
Learning highlights
- Learn ML System Design: work with real-world examples and create ML system design doc.
- Learn ML Engineering: work with popular open-source tools and create a production ML solution.
Course outline
15 classes
Intro
The difference between System Design and ML System Design. Understanding a business problem. Collecting requirements. Big services and small apps. Setting goals. How to measure success? Discussing examples.
Metrics
Metrics, Financial Impact. Offline and online metrics. Metrics pyramid. How to evaluate the financial effect of your project.
Solution Design
Collecting system requirements. Functional and non-functional requirements. Getting a top-level understanding of the solution architecture. Discussing examples.
Data Science Methodology
Formulating the ml problem, Feature Engineering, selection of ML metrics, selection of an ML algorithm, construction of validation. Discussing examples.
A/B Testing
Explore the purpose of A/B tests, key considerations for a successful launch, and the importance of validation using historical data. Understand scenarios where A/B testing is not feasible, gaining comprehensive skills to make data-driven decisions with confidence.
Deploying simple apps
Deploying a simple rule-based service in production (later you’ll replace rules with ML models). In the meantime, you’ll have a quick SE recap: VSCode, bash, Git, Poetry, ssh, scp, tmux, Docker, Docker-compose.
Easy UI for demo & testing
You’re going to learn Streamlit and Gradio and create simple UI applications that help demo your models, debug and test them.
Pretrained models FTW
Pretrained models are used everywhere right now, so let’s learn how to find the right one for your problem and how to put it in production, replacing our rule-based service. You’ll measure how accurate the pretrained model can be for our case and discuss ways of improving them.
LLMOps Essentials
Dive into the realm of LLMOps by using APIs or serving open-source models by your own. To test the capabilities of our classifier model and find examples for improvement, you’re going to deploy the bot on youare.bot platform and talk to him, wearing the white hat of a ML specialist.
Hackathon & Demo
You are going to talk to bots of each other via youare.bot platform and see who has the most accurate classifier, the most cunning bot, and who among us can distinguish bots from humans most accurately.
Mastering ML Experimentation
To train the model, you need to conduct many experiments and find what leads to quality improvement. You’ll explore the essentials of experiment tracking using MLflow.
HTTP web services
To make your model available to external people, you need to serialize it and build a service or pipeline that will apply it. Here you’ll explore serialization methods and scenarios of deploying models to production, including real-time serving with HTTP web services. You’ll learn about model deployment tools like FastAPI, TorchServe, Triton, and MLflow Models.
Model Deployment
Running your models on your laptop is good for demo, but isn't good practice for production scenarios. We’re going to deploy your service to fly.io or similar platform to make it available for your users even when your laptop battery is dead. We’ll also take a look at monitoring dashboards in Grafana and your application logs.
Hackathon
Time to combine all the parts of our solution together and test them in production. Once again, you are going to talk to bots of each other via youare.bot platform and see who has the most accurate classifier, the most cunning bot, and who among us can distinguish bots from humans most accurately.
Hackathon & Demo
The final test for our classifiers and bots. You’ll present what you did during these three weeks, and discuss the work on the project to find good practices and things to improve. We’ll also discuss how to make this project part of your portfolio, making these weeks beneficial not only for your skills, but also for your public profile and job opportunities.
Prerequisites
Machine Learning.
Deep Learning.
Software Engineering essentials: Git and Python, Docker.
Methodology
First week we’re going to learn about ML System Design and exercise to apply it to all kinds of applications, projects, and ML tasks you can encounter in your career. We’ll also create a System Design of the application we’ll be building for the 2nd and 3rd weeks. The 2nd week will be dedicated to deploying simple ML solutions and connecting them to the real system with real users. In the 3rd week, we are going to dive deeper into production ML by collecting the data, labeling it, training and deploying our own ML model. We’ll work on making our solution great in terms of model quality and robust in terms of service stability.
Grading
Alexander Guschin is an Industrial Head of Machine Learning at Central University and a Fullstack ML Engineer. During his career, he worked with ML in various domains and at different scales, both as an Individual Contributor and as a DS/ML team lead. He built companion bots with LLM and Generative models, contributed to the MLOps SaaS platforms and open-source MLOps tools, including https://dvc.org and https://mlem.ai. He worked as a Machine Learning Engineering Lead at a startup centred on the application of machine learning in the industrial sector and Data Science Lead in Yandex.Go. As a teacher, he co-authored the "How to Win a Data Science Competition" curriculum at Coursera, online MLOps class at Karpov.Courses and taught classes about ML competitions and Production ML at Data Mining in Action, the largest offline open data science course in Russia, with over 500 students each year.
See full profileAwards
- 1 place
McKinsey WomenHack event, 2021
Olga Filippova is a Lead Data Scientist at X5 Group, one of the largest retailers in Russia. She has over six years of experience applying machine learning models to address business challenges. Previously, Olga worked as a team lead in the banking sector, implemented machine learning solutions for manufacturing, and participated in the development of a library for monitoring machine learning models at the startup Evidently AI.
She is a co-author and instructor of various machine learning courses, covering topics such as data science competitions, MLOps, classical machine learning, machine learning system design, and machine learning for managers.
See full profileApply for this course
Building ML solutions: From System Design to Deployment
by Alexander Guschin, Olga Filippova
Total hours
45 Hours
Dates
Jun 30 - Jul 18, 2025
Fee for single course
€1500
Fee for degree students
€750
How to secure your spot
Complete the form below to kickstart your application
Schedule your Harbour.Space interview
If successful, get ready to join us on campus
FAQ
Will I receive a certificate after completion?
Yes. Upon completion of the course, you will receive a certificate signed by the director of the program your course belonged to.
Do I need a visa?
This depends on your case. Please check with the Spanish or Thai consulate in your country of residence about visa requirements. We will do our part to provide you with the necessary documents, such as the Certificate of Enrollment.
Can I get a discount?
Yes. The easiest way to enroll in a course at a discounted price is to register for multiple courses. Registering for multiple courses will reduce the cost per individual course. Please ask the Admissions Office for more information about the other kinds of discounts we offer and what you can do to receive one.