My final project - startup pitch generator


Background: I find most of startup mission or pitch are quite similar, is it possible we can create a startup pitch generator using Markov chain library provided by codecademy?

After more than 5 hours of work: [Edit: see profile for link]
I’m not pretty happy with the result, maybe it’s because the data is not enough for it to produce something meaningful. If you try to re-generate a few more times, you might have something interesting.

Data Source: At the beginning, I was trying to screenscrape AngleList, but it’s a pain to screenscrape their site because they are doing a pretty good job of preventing people to do that. I end up to use CrunchBase free API to get around 60,000 records.

Deployment: Unfortunately I would run out of memory on the Heroku free tier if I try to generate Markov chain on the server side, so I have to pre-generate the Markov chain and save it as a JSON file. That’s why the first time you load the site, it’s pretty slow, but once you have the JSON file downloaded, then it’s pretty fast to generate more string. The JSON file is host on Amazon S3

Main technologies used:

  1. Python flask - very light weight framework for hosting the site
  2. JavaScript - I have to generate string using the Markov chain on client side, so I rewrite the generate logic from Python in JavaScript

Please feel free to give me some suggestions. Here is the code: [Edit: see profile for link]