Putting Data-centric AI into Practice (Presented by Snorkel AI)
, Machine Learning Engineer, Snorkel AI
The current machine learning application development process is broken. Teams spend months, if not years, just collecting labeled training data. Then, they focus solely on iterating on the models, which has diminishing returns on model performance. Iterating on the data itself is the arbiter of success or failure in building machine learning applications. Join us as we explore this shift from a “model-centric” to a “data-centric” development process and techniques to speed up machine learning development by 10-100x. You’ll learn what effective data labeling and error analysis looks like so that you can iterate not just on your model, but also your data. Last, but not least, we’ll rethink model life cycle and explore how you can monitor and continuously improve your machine learning models.