2023-Japan-Einblick

PAUL YANG: Hello, everybody. Really excited to be here with you all today, to kind of show you a little bit of what we've been building. Einblick started really 10 years ago, when a MIT professor, who is an MIT professor now, started his postdoctoral research. But since then, we've kind of built ourselves into a solution that is intended to make data science effortless and collaborative.

It's a kind of a simple statement, but right now, if you have data scientists, data analysts, their day-to-day looks like this. It's coffee, unhappiness, and a lot of code on the screen. If you go and you go look at Google, 1 in 5 Google results about data science is about data science failing. And I think that's a fundamentally broken industry because it's like you would never read that about bridges. You don't go on Google, and 1 in 5 Google results is why are the bridge is failing.

And so data science is kind of struggling to find a place in a lot of organizations, because I think a lot of people are doing it wrong. They're doing it wrong because data science looks a lot like code, right, you write Python code. But it actually isn't very similar to code in the sense that it's a creative, exploratory analysis.

And so what we did was we first said, let's go and put data science on a visual canvas. Let's also change up the back end a little bit so that we can take in any number of languages. We can ingest from any number of sources.

When you open a Word document, you don't need to write five lines of code to open a Word document. So why do you need to write five unique lines of code to open any CSV file? That doesn't really make sense.

So we build a back end that kind of takes care of some of these things for you. And then put other things into place, like automatic machine learning, to make your life easier. There's things that you do as part of data analysis, that you do over and over again. And we can develop algorithms that kind of make your life easier and solve those for you. So AutoML will help you find the best machine learning model to solve your problem and help you be able to put that somewhere else.

So to give you the demo through video, even though we have a booth over there-- and please come visit us after this talk at lunch-- what we do is we connect it to a SQL data set. And here what we're doing is dragging out some columns. And you can see that it's very easy to visually explore this data set.

And I can see here that, oh, age actually has a weird distribution. So I actually now will write one line of Python code, to go ahead and clean up some of the data. And from there, the Python code then can be visualized as a table immediately.

And so, today, if you have data analysts who are working in some kind of tool, they're probably not experiencing a tool that looks as flexible and fluid as this. They're probably working in lines of code-- lines of code. And I call this the curse of software engineering, where data scientists were given tools, like VS Code or Jupyter Notebooks, that look a lot like the traditional IDEs of yesteryear, rather than something that's more creative and exploratory, like a Figma or a Miro, which maybe you guys are familiar with in other domains.

Here, we trained a quick machine learning model. We can get to explainers which tells you the top factors of the model. And here we see that the demo concludes.

So as an example of how we've actually been deployed in practice, data science, very flexible. All of you have data. All of you have more data than you're probably using today.

We work with this leading chemical manufacturer. And their question was kind of simple, when is the machine going to break? And they have sensors. They take more data readings than almost anything.

Every 15 seconds, they had 120 sensors reading results. And in the past five years, their machines had some breaking events. They had reduction of yield. And so they have a very clear, bad outcome they want to prevent. And they had 100 million rows of data every year.

And so they were able to take that data and put it into AWS as three buckets, connect to it directly. They worked together. And they developed this initial ML model, that was able to be produced in less than one week.

Previously, what they were doing is that the plant manager has 25 years of experience. And so he can hear the sounds. And he hears the sounds of the machine, the coalescer. And he says, oh, it's time to do a little bit of repair, or you can take a look at a lot of sensor readings.

So artificial intelligence is really just taking all those patterns that humans have been thinking about and looking at for the past 30 years, and turning it into a machine-based approach, to kind of make that decision, rather than listening to the machine and hearing it, being able to say, oh, well, sensors, like 5, 10, and 17, have this reading. We can make a prediction that this machine is going to break down.

We have existing relationships in Japan, both companies. And we also work with academics, even. We actually work with one academic, who's focused on English language learning in Japan. And a lot of these folks have more data than they know how to work with.

This guy is a professor in education. So he's not a data science expert, but he has data. He wants to do the data analytics. And so he's actually using Einblick here in Japan to work on some of his problems.

Our strength is in helping teams that are data-heavy, but maybe don't feel like you're using data fully yet. We want to partner with data scientists, business analysts. And we can do a pilot. And we can install on-prem, or we have also a free trial.

Please email me if you have any questions, py@einblick.ai. But, also, if you go to our website, which is einblick.ai, you can actually try for free.

So if any of you are interested in just having a look, exploring data on your own, all you have to do is give us an email. And it'll be ready for you. And we're at the booths. So please come see us.