MOPS Community

Hi Pugaholics!

We are pleased to invite you to the sixth edition of the MOPS meetup that is coming once again to Poznań!

The formula is similar to previous editions, about 30-minute practical talks followed by question-and-answer sessions with networking afterward.
The entire event and all talks will be held in English.

Key details:

• Location: Fandom office, 2nd Floor, Malta House, Baraniaka 6, Poznań
• Insightful talks 💬
• Knowledge exchange during networking with pizza 🍕

Plan of the meeting:
18:00-20:00 - Presentations:

1. Model serving 101 with FastAPI & LitServe (Marcin Zabłocki)
2. Ways to Scale Up ML Experiments on AWS (Julia Będziechowska)
3. Beyond the Hype: Leveraging possibilities of LLMs and Generative models in creating MLOps solutions (Piotr Kałużny)

20:00-22:00 - Pizza + Networking

Presentations:

• Model serving 101 with FastAPI & LitServe - In this presentation, I will take you through the journey of serving machine learning models—because a model unserved is like Schrödinger's cat, both useful and useless until deployed. We'll begin with an introduction to FastAPI, exploring why it's become a favorite among developers for its speed and simplicity, as well as discussing some of its shortcomings and quirks you might not see at first glance. Starting from setting up basic endpoints and handling requests and all the things you’ll need to take care of to serve ML models well. Finally, I will also show LitServe - a library that adds some ML-specific magic on top of FastAPI to ease the pain. Whether you're new to model serving or looking to optimize your workflow, this talk aims to equip you with practical insights and tools to effectively bring your models into production.
• Ways to Scale Up ML Experiments on AWS - Growing data organisations require agile approach to ML-powered experiments in order to gather feedback and iteratively improve the solution. On the other hand, dynamically changing landscape of the AI world is technically challenging to keep up with, in particular when considering big data volumes.
  In this talk, I will demonstrate how we leveraged various AWS services to successfully apply different Data Science projects at large-scale at Fandom. We will consider different layers of AWS solutions, from fundamental EC2s through serverless to Sagemaker - a service tailored specifically for ML - with tips & tricks, why we used them, and whether you should use them in your project!
• Beyond the Hype: Leveraging possibilities of LLMs and Generative models in creating MLOps solutions - LLMs and Generative AI are changing how standard ML process looks. They can redefine the processes we currently know as data annotation, classification, computer vision tasks and content creation. These novel uses, connected with concepts such as "LLM as a Judge" come with even greater need for stable MLOps pipelines. The needs vary from robust validation, specialized monitoring to handling and managing hundreds of prompts per case. In this presentation I will showcase how do I approach these problems while creating full-stack ML solutions.

Speakers:

• Marcin Zabłocki - MLOps @ Printify, youtuber @ ML-Workout.pl, likes to code and likes learning… machine learning.
• Julia Będziechowska - Julia works as Senior Machine Learning Engineer at Fandom and her everyday work focuses on turning AI prototypes into robust, large-scale solutions. She is also a member of the Poznań Chapter of WiMLDS (Women in Machine Learning & Data Science) organisation.
• Piotr Kałużny - Currently Senior ML Engineer at Allegro, former Data Scientist with strong analytical background developing ML solutions in the last 10 years. PhD specialised in applied-science ML in varying areas ranging from Telco, Financial Security to E-commerce. Old enough to understand and have used Data Mining approach or Hadoop, fresh enough to carry out projects utilising LLM and genAI models.