Do you want to have more impactful, profitable and predictable data science projects?
If you are a business leader who wants to better understand and de-risk your projects I can help.
If you're a data scientist who wants to understand how to deliver better, stress free results for your clients, I can help too.
How?
After working with and training thousands of data professionals at large enterprises, startups and through online training programs I developed, I started to see common mistakes and patterns.
Often data professionals can present and optimize performance metrics without fully considering the trade-offs and wider implications that ultimately determine the profitability of a data project or not.
There are plenty of reasons why a 99% accurate model can actually be bad.
Paper results don’t equal real world results but by knowing what to look for we can minimize that gap.
"If AI is a rocket, data is the fuel..." Andrew Ng
More data isn't always better though. More data may just mean more noise and more cost to store and process. Capturing the wrong data means your project will be doomed to failure.
Often this is not realized until sometimes years of investment have been made to build models but this doesn't need to be the case.
By developing a data strategy you can increase the profitability and predictability of your data science projects by understanding ahead of time what data is
- The right type
- The right quantity
- The most cost effective
Data strategy helps us to understand where the numbers used to train machine learning models actually align with the true business objectives.
It also helps us to understand how we can actively design data collection to improve model performance while reducing cost.
What can go wrong?
Overfitting, data leakage, data volume, data consistency, data quality, measurement error, non-representative samples, slippage, transaction costs, data timing, data interdependencies, restating units, lack of cost functions, unmaintainable code, survivorship bias, vanity metrics, inappropriate performance metrics, lack of a baseline and metric sensitivity are just some of the issues that can occur.
Just matching a data scientist with a load of data is not enough. Infact this can be an expensive and frustrating search for a needle in a haystack that might not even exist.
This is where I can help.
Sign up to my free newsletter to learn how you can start increasing profits and reducing risk with a solid data strategy today.
About Jonathan
Jonathan helps chief executives and their data science teams develop data strategy to produce more profitable and predictable data science projects.
His experience in business and technology allows him to design practical solutions that are more likely to fulfill the business needs.
He is a best selling instructor who has trained over 14,000 students and has worked with companies including HSBC, Morgan Stanley and leading tech startups on their data projects.
He has extensive experience of what real world data problems actually look like as well as a deep understanding of how machine learning models actually work.
Jonathan has been developing decision support systems since the age of 15 when he started studying computer science, information systems and business at The University of Auckland. He has over 20 years of commercial experience working with data and technology.
He is passionate about helping others to implement technology that will makes a positivie difference to the world.