MindsDB: Automated Machine Learning with Jorge Torres

Automatic TRANSCRIPT

Or. Hey welcome to the show. I geoffrey thank you for having me. You work on minds db. What problem is minds db trying to solve. We're trying to solve the problem of making it very easy for you to apply machine learning if your data is in a database. Give me a little bit more context wise. Is that a problem that is worth solving the problem that we see people have when they have a prediction that they went to make is that they didn't to reinvent the wheel over and over and maybe the best way to cover business and example. Imagine that you have a database with inventory information and you want to forecast how your inventory listing in the next week or the next day next month and traditionally how you go about. This problem is that you have davis scientists or a machine learning engineer that goes into your database extracts inventory for your iphone mentor and then loads into a data frame panzer frame. They usually work on fifa notebook and they go and build a model spent some time weeks. Maybe they really good build a model and then once they helped this model day. Even a model is pretty good. They run into a wall when they want to move this model into production. What i mean by this is usually in that same database. You not just have an iphone. For many of the stories you may have thousands of products ten thousand products or even you may have tens of thousands of products as well as many stores that you may have and what that means is that you're going to have to train tens of thousands of models for solving that particular problem off predict inventory. Which is unviable like. There's no way that you're going to train a model for each of the products that you have on each of the stores that you have so given that. You're davis rain. The database the workflow that minds be enables for people to say well. Actually i want to predict this call from this table or from this query and figure out the rest and what does is that allows you to do two things. Tell the database itself with simple command. Like i want to make this prediction. Usually do this with an insert statements and the to be worlds and you do it straight into database and then what might be behind the scenes says that it figures out. What's the best way to train a model. That is a time series model for this task and then it will publish just model as a table that also lives in the database and then you can query predictor the same way that you create a table so you can do okay now from my select from predictive model where the product is book. And the data's tomorrow and it will tell you the inventory value from

Coming up next