MOPS - Meetup #2

Hi, Pugaholics!

We are delighted to announce the second edition of the MOPS meetup is finally here! Once again, it will consist of three 25-minute practical talks, followed by a 5-minute question-and-answer session. The entire event and all talks will be held in English!

Key details:
Location: Allegro office, 8th floor of Fabryka Norblina in Warsaw
Perk: Free talks & pizza, knowledge exchange during networking

Plan of the meeting:
18:10-18:40 Portable and cloud-agnostic MLOps Platform with Kedro, MLflow and Terraform (Marek Wiewiórka)
18:40-19:20 Architecting ML for Huge Impact and Scale (Szymon Jacoń)
19:20-20:00 Building machine learning platforms for a living - lessons learned (Mateusz Kwaśniak)
20:00-22:00 Pizza + Networking

Presentations:
Portable and cloud-agnostic MLOps Platform with Kedro, MLflow and Terraform - The selection of managed and cloud-native machine learning services that you can run your data science pipelines and deploy your trained models is versatile. Unfortunately, there is no single way of interacting with platforms like Amazon Sagemaker Pipelines, Google Vertex AI Pipelines, Microsoft AzureML Pipelines, Kubeflow Pipelines – or recently SQL and Snowpark within Snowflake data warehouse. In this presentation you will learn how a production-grade MLOps Platform built by GetInData and powered by battle-tested, cloud-native and/or open-source technologies such as Kedro, MLflow, and Terraform would make your data scientists’ life easier and more productive - regardless of what cloud provider you use.

Architecting ML for Huge Impact and Scale - In 2022 Allegro Pay financed over 5B PLN worth of orders on Allegro, for more than 1M customers while maintaining low credit risk and stellar NPS. A decent portion of these results was driven by ml models deployed at scale. We will kick off with describing how raw events are mapped to analytical entities on top of which feature engineering happens. This, along with our plug-and-play experimentation pipeline, lays foundation for effortless model training. All experiments are registered in MLFlow, and the models stored there can be deployed in either batch or real-time mode. Finally, we will showcase our decision engine, the ‘brain’ behind all Allegro Pay’s real-time data-driven processes, and how it leverages model predictions and the feature store. Veering on the technology side of things, there will be key takeaways about tools and frameworks currently in use: Airflow, dbt, MLFlow, Snowflake and numerous Azure components.

Building machine learning platforms for a living - lessons learned - Building MLOps ecosystem and platforms is still a relatively new and challenging topic for many companies. While each case might look very different at first, there are some interesting, repeatable observations, patterns and issues that many ML Platform teams struggle with.
This presentation delves into the experiences and insights gained from designing, building, and operating ML platforms in production environment. We are going to look at the real-world examples and practical tips to help the audience navigate the complex landscape of ML platform development and stakeholders management.
Whether you are a data scientist, engineer, or manager, this session will provide you valuable lessons and takeaways to enhance your ability to build and manage successful ML (or MLOps) platforms and ecosystems.

Prelegents:
Marek Wiewiórka - Marek is a seasoned Big Data and Cloud Architect with 15+ years of experience in designing and implementing modern data and MLOps platforms. Currently Chief Data Architect @GetInData | Part of Xebia and Research Assistant at WUT putting the finishing touches to his Phd dissertation. In the past, a data engineer/consultant at Truecaller, ING, PLAY, Volt.io and several more companies. Privately - a keen long distance runner, gravel bikes enthusiast and absolutely in love with the Italian Lakes and Strade Bianche of Tuscany!

Szymon Jacoń - works as a Junior Machine Learning Engineer at Allegro Pay since March 2022. My job involves improving our ML platform and putting ML models into production. I earned a Bachelor of Engineering degree from TUL. In my free time, I’m enthusiastic about Python and machine learning, always eager to learn more.

Mateusz Kwaśniak - I started my professional career as a Software Engineer, then I got attracted to data science and machine learning. At first I wanted to switch to a Data Scientist role but I quickly realized that I should rather stick to the engineering side of the projects.
Now I combine my engineering expertise and passion for machine learning at deepsense.ai where as a Lead MLOps Engineer I actively participate in design and development of MLOps platforms and tools for our enterprise clients.

Find us on Linkedin