Surgical ML Model Debugging with Sliceline: Find the Blind Spot!
Details
When operationalizing ML models, one often underlooked crucial risk is having your model underperform on some data slices. This effect could have disastrous consequences for your organization. In this talk, Antoine De Daran will present Sliceline, an approach to automatically and efficiently find data slices where your model is underperforming, both when validating and monitoring. The methodology is based on a research paper from Graz University of Technology and an open-source Python implementation by DataDome.
⏰ Schedule
- 18h30 - 19h00: Welcome & Introduction 👋
- 19h00 - 19h45: Talk 🎙
- 19h45 - 20h00: Q&A Session 🙋♀️
- 20h00 - 21h30: Cocktail & Networking 🥂
🎙 About the speaker
Antoine De Daran has been a Data Scientist at DataDome for the past 2 years working on anomaly detection projects to automatically identify bots in web traffic. Previously, he worked in consulting firms as both a software engineer and a data scientist.
🙌 Sponsors
DataDome is a fast-growing tech scale-up, leading the bot protection industry from New York, Paris, and Singapore. DataDome’s online fraud & bot management solution offers full protection—across mobile applications, websites, and APIs—in real time against online fraud, including web scraping, account takeover, layer 7 DDoS, and payment fraud.We’re proud to protect more than 200 clients, including TripAdvisor, The New York Times, Footlocker, Hanes Brands, Carrefour, BlaBlaCar, Rakuten, Veepee, Adevinta and many others.
Ekimetrics, founded in 2006, is a European leader in 'data science for business'. The company’s mission is to help customers audit their data opportunities, enrich their analytical capital and deploy actionable solutions to maximise their marketing and operational performance and re-energise business models. With more than 320 data scientists, Ekimetrics is one of the largest independent teams in Europe. Ekimetrics' focus is delivering short-term gains while ensuring the long-term development of its customers' data assets.
nibble is a technology player in the MLOps space helping organisations scale their data-heavy algorithmic initiatives (AI/ML and the likes) with a mix of service and in-house software solutions. Their main product is a Feature Platform called spice.
🤩 Audience
This is a meetup for data professionals who want to push forward the productionization of machine learning development within their organizations: data scientists, data engineers, software and devops engineers, data project/product managers...
👋 Contact
For any inquiries, please contact florent@mlops.paris