New Project | Prediction Engine

Sooooooo this is usually kinda like a boilerplate portfolio “learn to code” kinda project but I want to take it a few steps farther and apply it to something useful.

You know when you use Netflix and it recommends new movies to you? That’s a prediction engine.

Now these can get crazy complex but I really loved working with Vector Similarity Search which is what I want to start off basing mine on. I might get more crazy in the future.

Vector Similarity Search (or lazy algorithm or k-nearest) is kinda like the lazy man’s machine learning. And an easy way to “do AI”. Think of it like we have a personality test measuring Extroversion/Introversion and Smart/Dumb.

Everyone gets a score like a point on the graph. A smart extrovert may be: (2, 3)

And a dumb extrovert might be (1, -2)

Draw these as points on a graph and if you put yourself on the graph, you could find the closest K points and those would be the results of your search. You just found the K number of people most similar to you based on these characteristics.

So that’s how we’ll find the best recommendations. But I’m not gonna do it with 5 fake movie titles. I’m going to use a real and practical dataset so I’ll have to find a way to make this efficient, ideally on my own instead of just chucking langchain on it.

I’m going to use Government Contracts. If you go to sam.gov you can browse government contracts up for bid. And as a small business you may want my lovely service to recommend bids for you based on the ones you like!

So join me as I work through this bit by bit

I’m also putting this on Twitter here:
https://twitter.com/SolContractADay