Building Data Science Products - Think Business First

Explore a comprehensive 31-minute talk on the importance of prioritizing business considerations when developing data science products. Learn why focusing solely on tools and techniques can be detrimental, and discover a high-level thinking process for conceiving, implementing, deploying, and maintaining machine learning systems. Gain insights into asking the right business questions, choosing appropriate algorithms, and understanding the machine learning model life cycle. Delve into crucial aspects such as data quality, metrics selection, repeatability, business impact, and the 80/20 rule. Understand why data often trumps algorithms and how to evaluate model predictability. This talk from Data Science Dojo challenges common myths and emphasizes the significance of a holistic approach to building successful data science products.

Intro
About the Speaker
Why This Talk?
Is a Highly Accurate Model Always Desirable?
The 'Data' Question
The 'Metric' Question
The 'Repeatability Question
The 'Business' Question
The 'Impact' Question
Modern ML Tools - A Blessing or a Curse?
Machine Learning Model Life Cycle
What is your biggest pain point?
Data beats algorithm
Dispelling a Common Myth
The 80/20 rule
Beg, borrow and steal
Building the model is only a small piece of the puzzle
The Importance of Data
Predictability of your model
Further Evaluation Data Set
Increasing the exposure of your model