MOPS - Meetup #3

Long time no see Pugaholics!

Again, we have the pleasure to invite for the MOPS meetup!
The formula is similar to previous editions three, 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: Allegro office, 8th floor of Fabryka Norblina in Warsaw (entrance Plater 3, from Żelazna Street) 🏢
Free talks 💬
Knowledge exchange during networking with pizza 🍕

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

1. Scalable, secure, and sustainable NatWest Group’s MLOps platform (Marek Brynda)
2. MLflow iceberg: from basics to hidden depths (Marcin Zabłocki)
3. Building a vector search engine (Maciej Mościcki)

20:00-22:00 - Pizza + Networking

Presentations:

• Scalable, secure, and sustainable NatWest Group’s MLOps platform - Presentation covering Natwest Group's MLOps platform, starting with problems to overcome, ideas how they were solved and ending up with implementation - focusing on Data Scientist/ML Engineer points of view.
• MLflow iceberg: from basics to hidden depths - In this presentation, I will guide you on a journey through MLflow, a popular platform among MLOps practitioners. Just like an iceberg, MLflow has its visible features and a lot of powerful tools and gotchas below the surface. Starting with the basics of metrics logging, experiments, and comparison, I will gradually delve deeper, revealing the more nuanced aspects of deployment options and their potential challenges as well as customization options and recently added LLM features.
• Building a vector search engine - Vector search is ubiquitous - powering applications such as visual search, semantic search or more recently Retrieval Augmented Generation. As MLops engineers building vector search solutions, we need to make decisions which search engine to use. While there are many off-the-shelf products on the market I would like to offer an alternative approach. During the presentation we will discuss how can we build a vector search engine ourselves using FAISS - a popular library for similarity search. Along the way we will learn about algorithms for making the search faster and more resource efficient. I will also share tips about deployment and measuring the quality of our search engine.

Speakers:

• Marek Brynda - Data Scientist at NatWest, with a background in physics that I found less appealing than expected. On a daily basis, I experiment with new approaches using classical ML models, fit model code to SageMaker templates, and analyse results. Beyond the data realm, I have 3 fluffy cats
• Marcin Zabłocki - Marcin is an explorer of new technologies, not only in ML. He likes to quickly deliver working software where no obvious solution is known. He started as a full-stack web dev with a passion for C#, then Data Engineer, ML Engineer, and now MLOps Architect. Familiar with every piece of technological pie needed in ML projects. He likes to share his knowledge in all ML-related areas with a soft spot for NLP (he started in the TF-IDF era, years before Chat-GPT became cool). Kedro community activist and member of TSC.
• Maciej Mościcki - Senior Software Engineer at Allegro. For several years he worked on introducing various ML solutions - such as neural network based ranking and semantic search - to Allegro search engine. He was one of the founders of Allegro MLops platform. Nowadays he is primarily focused on internal technical platform - async communication, authorization and deployments system.

Find us on Linkedin