Data Science at Fiverr: From Research to Production

Join us on a special evening where you'll be able to connect with like-minded individuals, learn from our Data Scientists at Fiverr, and explore the latest trends and technologies in data science.

This meetup will be divided into four main parts: research, data, modeling and production.

Mark your calendars and get ready to immerse yourself in a world of data-driven insights and innovations at Fiverr's data science event!
**This event will be held in Hebrew**

AGENDA:

18:00-18:45- Sign up, mingling, shmingling and some delicious food

19:00- 19:20- Research: Shifting Fiverr Search into Deep Learning/ Gal Naamani

19:20-19:40- Data: Detecting Suspicious Purchases on Fiverr/ Idan Lenchner

19:40-19:50- 10 minutes break

19:50-20:10- Modeling: Content Moderation at Fiverr/ Tomer Ron

20:10-20:30- Production: Promoted Ads in Search Results/ Tom Cohen

DETAILS:
Research: Shifting Fiverr Search into Deep Learning, Gal Naamani
Search is one of the biggest contributors to Fiverr's revenue. As such it is home to many of our machine learning algorithms, one of which is an advanced ranking model responsible for choosing the order of which gigs are shown in a page. In this lecture we'll talk about research done before deciding to shift the model into deep learning. We'll describe the methodology of how we answered important implementation questions even before developing any code.

Data: Detecting Suspicious Purchases on Fiverr, Idan Lenchner
In this lecture, we'll delve into the world of data and explore its crucial role in shaping AI models.
Our talk will focus on the critical importance of data quality, we’ll explore different EDA methods as well as strategies for selecting the right features to optimize your model's performance.

Modeling: Content Moderation at Fiverr, Tomer Ron
Fiverr heavily relies on user-generated content (UGC). In this lecture, we will discuss how AI models are integrated into Fiverr's content moderation flow. We will explore the model architecture we chose and explain our decision to use an attention layer. Lastly, we will talk about model auto actions and how we evaluate it.

Production: Promoted Ads in Search Results, Tom Cohen
During this talk, we'll cover the data science flow of a bidding model and discuss the stages involved in creating a successful model at Fiverr. We'll focus our session in the best practices for deploying models in production, the importance of a feature store in a machine learning pipeline, and how it helps us to streamline the development process and the production serving.

**This event will be held in Hebrew**
**Entry upon presentation of ID**

We look forward to seeing you,
The Fiverr Team.