I…… I what? AiSara

Two years ago, I was staring at my screen, trying to analyse what seemed to be a complex problem. The details are not important, but I told myself there must be a solution to it. Apparently not!

Fast forward to the present time we are proud to release AiSara to the market. In total, we probably have spent about 8000 working hours (5 man-years) on this. When I said we, it is the hardcore team consisting of a “let me have a crack at this” engineer, “never say die” developer, and many others (mostly interns) that were unfortunate enough to cross our path. My only prayer is that what didn’t kill them made them stronger, lol.  

When people ask me what AiSara is, I figured there is no better way to answer than to show what it is. Initially, it was like trying to explain what a touch start button for a car is 20 years ago, except it is 10 times more difficult. What usually follows is a perplexing look, people guessing what they are or even suggesting how it should be. 

What is more important is what AiSara does. Earlier this year, after 1 year of working on the algorithm and 1 year developing a web app that many possibly have a mental block to, we figured we need to release AiSara as an Excel add-in. With a user base of 750 million around the world, at least we don’t have to explain the platform, and that is already half the battle won.

What is the other half of the battle? Again, with Excel, it is much easier. With AiSara, you can boil it down to two functions we created for Excel.

=LEARN (input_range, output_range)

In essence, with the above function, you give AiSara your dataset, and you tell it what the input variables are and what is/are the output variables.

Why? You guess it, so we can make a prediction. Next, after teaching AiSara the relationship between the input and output from LEARN above, we can predict as follows

=PREDICT (the stuff that was taught above represented by a cell in excel, input parameters, and which output you desire in case there is more than one)

Isn’t it simple? No coding, no struggling with thinking about neural network models. Wait, if it is that simple, why wasn’t it available before? I remember something I read in the book Simplify by Richard Koch & Greg Lockwood, to create something simple is not simple! (Example: iPhone). Our first research for AiSara was how do we translate the way humans see a pattern to an algorithm. I think you can imagine what a challenge that can be. The end result is an algorithm that can make a reasonably good prediction even when a large dataset is missing, as long as it can see the existing pattern. The image below shows AiSara filling up the blanks in the middle accurately, with no human intervention.

Sinusodial Plot

I invite you to explore AiSara and see how it can be used in your optimisation problem, diagnostics and predictive analytics or just check it out to see what the fuss is all about. We have made it easy firstly for us, and if it easy for us we think it will be easy for you. 

So, we have climbed Kilimanjaro, where do we go from here? Tell us here what to fix, and what to build! We can work on unsupervised learning, focus more on classification versus regression, anomaly detection, big data, enterprise version, and many more areas, but we like you to influence us to build a product you will love. 

As Uncle Ben said to Peter Parker, “With great powers come great responsibility!”. We predict there will be good demand for a strong yet universal “plug and play” predictive power. Therefore, in the works is a stock market prediction app, (also as an Excel plug-in) using AiSara engine as a demonstration of AiSara Enabled app. Watch this space!

AMZN Blindtest

Our vision is to see this logo proliferate around us and power a lot of predictive analysis software, services. We realise we are still a long way from there, but the day that happens, we know we have made some impact.



On behalf of the AiSara Team,

Zaim Awang.