Implementation

Implementation

The project has multiple components and is implemented using well-known cloud applications.

At the heart is a python script. This is where I make API calls to TD Ameritrade to pull quotes and send orders. I also have a number of checks to ensure I have sufficient funds, my account is in good standing, etc. When important actions happen (e.g. an order is sent, a call is assigned to me, or importantly, if any part of the script fails), the script sends me an email alert via Mailgun with the details. The script runs daily on a simple Heroku dyno (this is a lightweight Linux container that lives in the cloud. I use Heroku so that the whole program can run without my computer.

In my predictive model there is a heavy-duty calculation that requires a lot of RAM, so I use an AWS Lambda function to compute predictions. The python script sends a request with input data and the Lambda function sends back the predicted distribution.

In order to store data, predictions, and anything else that I want to persist after the script finishes executing, I use an AWS S3 bucket.

There is a second script on Heroku that pulls the data stored in S3 and creates a plot comparing my prediction from 10 days ago to the actual outcome. It sends that plot to this site (see it here).

All together, it looks like this:

You can see the code (and a few other projects) on my GitHub