Data Operations IL: Challenges of Data Collection
What is it that makes data collection so challenging and full of hurdles when it’s so extremely crucial? It’s the first step of every data-based project, and without it you don’t even get to say “garbage in garbage out”.
Data Operations IL & UVeye have joined forces for this event to introduce and discuss the world of Data Collection - where data is harvested in the wild and sorted to fit business logic. This is the foundational step that powers many data pipelines and initiatives, and it’s far from easy and straightforward.
Data collection is one of the widest ranging challenges of data operations - across industries, everyone needs to get their data somewhere. Some are lucky and their data gets handed to them on a silver platter. Others, unfortunately, are not as lucky - and to them we dedicate this meetup.
We’ll hear about the challenges of some real world examples - computer vision for vehicle inspection and damage detection, AI solutions for high-stakes people analytics (e.g. hiring), AI image enhancement and visual upscaling- we’ll explore the methods, challenges and hurdles on the way to collecting diverse and high quality data for AI and ML models.
We’ll hear 3 interesting talks from seasoned data collection professionals:
Dr. Kristina Zaides & Inbal Friedlander, Data Operations Team Leads @ UVeye
In this talk we'll hear about the advanced labeling challenges in automotive computer vision systems - dealing with edge cases, balancing existing standards and perpetual improvement, and how to manage huge amounts of data while maintaining consistency and effectiveness.
Dr. Adi Sarid, CEO @ Sarid Research
In this talk we'll hear about the challenges of data collection and data operations for predictive AI analytics in elite technology program recruitment process. The collection process here is very varied and changed (by design) throughout the process - with the goal being finding human talent in a fair and accurate manner.
Tom Shai, Data Operations Manager @ Microsoft
In this talk we'll hear about the challenges of collecting and labeling data in the Microsoft scale - from automatic recordings, smart labeling with neural networks, and advanced analytics on metadata. All this while keeping high regulatory and compliance standards, with the goal of training models to improve image and video quality for end-users.
*Talks will be given in Hebrew
Schedule:
18:00 Mingling & Snacks
18:30 Opening Words
18:40 Talks
20:00 End of Event