Data Collection
Baebot: A SMS marketing chat-bot that makes sure my girlfriend is happy today:D
Built on Twilio’s API, baebot was written with Big Hero 6’s primary-care home robot “Baymax” in mind. With 2020 starting off as a complete shit-show, I wanted to make sure the ones around me were happy and if not, had a quick easy way to let someone know.
We all want to stay connected and have uninterrupted communication, but jobs, education, or personal work can sometimes get in the way. The bigger picture of this project is going to sleep knowing that the mental health of friends and family are okay.
This project is under Data Collection since I’m able to grab a CSV of all entries and potentially create a model. Eventually, when I have enough data to create a model for measurements, I hope to move this project under Predictive Models. I want to show this data’s correlation with weather, work-load (such as month end for accounting), and other factors I think of on the way.
SmartX: A StockX Projection Model
SmartX utilizes data collection to create datasets for a time-series forecasting model. Since this data isn’t available through a public API, this required an unofficial-API approach which allows me to grab every transaction made on an item, along with the details I need to construct the dataset.
Additional web-crawling in this project:
Grabbing SKU IDs, a needed parameter for API use since each size has its own.
Crawling blog posts for seller-fee promotion dates. These dates are significant since prices can fluctuate on “No-Fee” days. Incorporating them into the Holidays module lets the model see if those have an impact on price.
Details of this project can be found under “Predictive Models”
SmokeyTheBear is one the first programs I wrote and is still used, but not so much recently because Covid-19 and California fires:(
I’ve always been a huge fan of camping and so have my friends and family, but the first problem of it is getting a reservation. Since many people book so far in advance, it became harder to plan and make spontaneous trips. With my constant refreshing, something I noticed was people cancelling their reservations the week of, but it would happen randomly and often be swooped up by someone else.
This inspired me to write a monitor for reservation availability. Simple request monitoring and JSON handling took care of the data handling, while TwitterAPI was used for notifications via tweets and date-changes were handled with Direct Messages. Version 1 worked great for a few years. Friends and family were invited to use it for any of the 280 California State Parks supported by the reservation website and everyone loved it.
In late 2019, the reservation website changed their entire backend, rendering all my code kind of useless. But I saw it as a new challenge and dove in right as I saw the changes. This time I wanted to take a new approach and write it in an asynchronous format, which would allow me to build it with Discord’s Bot API.
Doing so made it SO much more efficient, allowing for automatic reservation booking with reCaptcha handling, easy date changes, adjusting delays, multiple date monitoring, weather checks, and all with the ease of Discord’s permissions handling. Another plus was their new backend-API, which after a ton of trial and error, gave access to so much more data on camping-spot-specific details like shade, campfire rings, grills, benches, and photos of the spot.
A nice cherry on-top was finding out that other states also the same backend for their state parks ;)
Helga Patak-AI
Helga G. PatakAI is another machine learning project, created after the Interview, to generate poetry based on the Hey Arnold! character “Helga Pataki.” Throughout the series, Helga recites what I think is amazing poetry for a 90s kid show, all about Arnold whom she’s insanely in love with. The show ended in 2004, but I still watch it all the time for the nostalgia.
When Covid-19 began causing lockdowns in March-April, Craig Bartlett (the writer and creator of the show) starting posting drawings of Helga and Arnold wearing masks in lockdown and writing to each other by mail. This really inspired me to create something that could bring a piece of everyone’s childhood back just as Craig did.
This was around the same time I was re-doing the Interviewer with GTP-2 so I immediately knew I could build one for Helga’s poetry. A huge obstacle I ran into was getting a dataset of all her poems. The Hey Arnold! Wiki Fan Page had a few, but not all. It took a few days, but I watched all five seasons and manually logged each and every poem lol! I definitely will be sharing this with the Wiki page so everyone can enjoy the collection.
This is a work in-progress; hopefully moved under Predictive Models soon;)
