Applied Machine Learning at Scale // a meetup by Pipedrive and PyData

On Wednesday, September 4th, Pipedrive hosts Applied Machine Learning at Scale meetup in Tallinn, co-organised with PyData Community in Tallinn.

When? 04.09, 18:00
Where? Pipedrive Tower, Mustamäe tee 3a

Agenda:
18:00 - Doors are open🚪; food and drinks ready 🍕
18:30-20:30 - Keynote speakers 🧑‍🏫
21:00 onward - Networking 🗣️; more food and drinks 🍻

This event will be held in English. Due to limited space, we’ll offer spots on a first-come, first-served basis, so sign up fast to secure yours!

Idea authors and co-organisers:
Andreas Beger, Data Scientist.
Isaac Chung, Senior Research Engineer.

Talks:

1️⃣ 22k models for the Engagement Score: scalability and performance

Denys Kolomiiets, Data Scientist at Pipedrive
We'll discuss the approach of training 22k models for different companies on a single machine. We'll explore how parallelization techniques allowed us to train these models simultaneously without the need for complex infrastructure or dividing tasks across multiple machines. This method streamlined the process, while significantly reducing operational complexity and allowing to train models on a daily basis for minimal cost.

2️⃣ Fraud attack prevention: Anomaly Detection Algorithms in practice

Elizaveta Lebedeva, Senior Data Scientist at Bolt.
At Bolt's fraud team, we track numerous time-series metrics to keep fraud losses under control. Our goal is to detect fraud anomalies as quickly as possible, preventing them from escalating into serious issues like fraud attacks or payment outages. In my talk, I'll share how we built and implemented anomaly detection algorithms to safeguard our systems, including the journey of designing, developing, and fine-tuning these algorithms while tackling multi-dimensional data challenges and reducing alert fatigue.

🎤 Meet your hosts

Maie-Liisa Sildnik, Senior PR Manager at Pipedrive