What I have learned working with scientists in my time at Amazon about building algorithms for solving business problems.

For the past few years I have been working with Amazon’s Machine Learning University. My role has been to help them scale and reach more Amazonian’s. We accomplished this by creating a learning platform with supportive course content, an emphasis on strong visual and interactive learning. We basically flipped the classroom. My focus has always been to put the experience of the learner first. Providing a clean design, removing noise, with engaging content so the student can focus on learning. The approach has been a great success in teaching Amazonian’s to solve Amazon business problems with Machine Learning.

Enough on what I did and back to what I have learned.

I think one of the first lessons was the importance of gathering data. Data is at the core of understanding the problem. We often approach problems in business or personal for that matter without taking in good data. We lack a complete understanding of the problem.

Until you have data you really can’t build an understanding of what ground truth is. This is very important because everything builds from ground truth. You must explore and visualize your data to establish this. You need to make sure you understand the details that may be hidden in your data. Establish things like context, sentiment, frequency to establish relevancy. Remove the unimportant stuff and focus on what is really important. To do this you may have to do the math, yes, the math. Math is the magic used manipulate and visualize your data. This may sound simple but remember its math. There may be some work to do as you build your algorithm and try and find ground truth. You may need to do some tuning, adjusting your algorithm there is not always a predefined path but more of a process or framework to help you get to your final output.

It is at this point, the point when you see things clearly. But that’s not it. There is something bigger than solving the problem. What? As you know solving problems can sometimes be like putting your finder on the dike. The problem may come again or one like it.

The real value comes in prediction. Now that you can see clearly what has happened in the past and today using insights from your data you can start to predict what will happen next. It’s something Amazon does very well. You know “I see that you purchased that guitar, would you like to look at guitar for dummies or perhaps something to tune that thing please. It’s the secrete sauce, yes, I’d like to solve my problems in the future.

There is one more thing and it maybe my favorite. In my time at Amazon I have discusses quantum physics, robots & rocket science and lots of other geeky fun topics sometimes while playing ping pong. What a privilege that has been to work with some truly inspiring people at Amazon. That said the real insights I gained where from working with the amazing people at Amazon. It’s their drive to solve problems, innovate, willingness to work hard and learn as they go. I’ve worked with some of the smartest people I have ever met here. Many of them are physicists, mathematics masters and rocket scientists. Yet they don’t pretend to know the all answers but they do know how to work hard to get them. I also appreciated the great leadership I worked with in my time at Amazon. Nothing happens in an organization without true leaders and inspiring people who establish clear expectations and remove obstacles it’s truly part of the algorithm for success.

Jeff

Instructional and Learning Experience Design

Jeff@mojocat.com