Recommendations leveraging Picnic's ML Platform & Data-Centric AI with Cleanlab

Join us for the upcoming PyData Amsterdam meetup that we host in collaboration with Picnic.

We will delve into Picnic's state-of-the-art tech stack for ML - aiming to empower engineers and enhance customer experiences and we'll learn more about Picnic's recipe recommender. In the 2nd talk, a paradigm shift in AI thinking will be discussed: from model-centric towards data-centric approach. We'll also learn how to improve your data quality with the open source package Cleanlab. Sign up for an evening of groundbreaking insights, discussions and networking!

Schedule
18.00-19.00: Walk in with drinks and food (🍕 /🍺)
19.00-19.45: “What’s for dinner? Building an ML Platform to enhance online-grocery shopping at Picnic”
19.45-20:00: short break
20.00-20:45: “Automated Data Quality at Scale using Cleanlab”
20.45-22.00: Networking + drinks and bites

[Talk 1] What’s for dinner? Building an ML Platform to enhance online-grocery shopping at Picnic.
At Picnic Technologies, we’re revolutionizing the way people buy groceries. Our affordable and sustainable service is made possible by cutting-edge technology and passionate engineers. Machine learning is at the heart of many of our operations. From time-series forecasting for the supply chain to recommender systems for our customer app.

In this presentation, Tom Steenbergen, ML Platform Lead, and Giorgia Tandoi, ML Engineer, will dive into Picnic’s machine learning platform and one of our ML use cases: recipe recommendations.

First, Tom Steenbergen will illustrate Picnic’s tech stack for machine learning and how it empowers ML engineers to easily run tens of models in production. Following the platform overview, Giorgia Tandoi will take the stage to discuss a quintessential Picnic project: recipe recommendations. This segment will explore how Picnic utilizes its technology to enhance customer experience by helping them discover delightful meals.

[Talk 2] Automated Data Quality at Scale using Cleanlab
Machine learning has traditionally followed a model-centric approach, with the primary focus placed on optimizing model architecture, algorithms, and hyperparameters while treating training data as largely fixed. Take GPT models as an example. From GPT-1 to GPT-4, OpenAI researchers found that simply training on “more data” did, in fact, improve performance. That paradigm, however, is coming to an end, as performance increases plateau with just more data.

A book from 2016 explored why small data is the new big data. While perhaps ahead of its time, the ideas were prescient for the challenges AI practitioners face today. With massive datasets readily available, it's no longer sufficient to just collect more data. Rather, the focus must shift to creating more high-quality, representative data. This revelation is fueling a paradigm shift towards data-centric AI.

In this talk, Aravind Putrevu will talk about:

• What Is Data-Centric AI and Why It Matters
• Limitations of the Model-Centric Approach
• Data-centric AI in LLMs
• Data Quality with Cleanlab

Final note
For our international PyData enthusiasts who are unfortunately unable to join us face to face, and participate in networking and discussions, you can still join us online via this link: [https://teampicnic.zoom.us/j/84038721932?pwd=cCtCMnNYSE9vbEROd0pweXlMYmg0UT09](https://www.google.com/url?q=https://teampicnic.zoom.us/j/84038721932?pwd%3DcCtCMnNYSE9vbEROd0pweXlMYmg0UT09&sa=D&source=calendar&ust=1706430842278065&usg=AOvVaw0TOsVDq6naO4tmopZfG_Os)