Grad school

October 2018

My primary career goal for a long time has been to be a successful startup founder. At one time, I was planning to get a part-time job after graduating (working only enough hours to pay for living expenses) so I could devote most of my time to building a startup. For a few reasons I decided not to try that, so I accepted a full-time offer from Lucid in order to save up money and gain experience. So my plan has always been to work here temporarily and then work on a startup full-time as soon as I'm done.

Over the past month or so as I've spent more of my spare time working on projects in Clojure, I realized also that there's a lot of stuff in the Clojure ecosystem I'd like to get the hang of. I started to modify my plan slightly to leave some space between quitting Lucid and doing the startup so I could spend time learning various Clojure libraries.

But in fact, the more I've thought about it, the more I want to emphasize this period of just learning. I've started euphemistically referring to it as "going to grad school" because I want to have a significant higher-learning experience. I don't want to go back to the structure of formal education though, so an actual master's degree wouln't hit the spot, and a PhD is out of the question because

  1. I found out in my undergrad that I don't like trying to fit my projects in the category of "research"

  2. I don't want to put off the startup for another 4+ years

So instead I'll just stick with my do-it-yourself, homeschool-style mini-master's degree from the graduate school of programming at O'Bryant University. It's not accredited, but I know the headmaster and I think he's great.

In preparation, I started listing subjects to learn more about and included a few projects that I've wanted to code up for a while. I realized that the projects actually were great ways to put into practice all the things I wanted to learn, so a natural structure for my curriculum emerged. Here are the projects.

Project 0: Write an AI to play Clue

Whenever I play Clue, I feel like there's so much information and I can only latch on to part of it. A program could keep track of every little detail. The main part of the program (i.e. given the information you have, which cards are in the envelope?) would be a good opportunity to practice logic programming, but there are other components too. Which room should you travel to? What questions should you ask? I tried to build this in my Intro to AI class but the professor wouldn't let me because our final project was supposed to involve robots.

Project 1: Write a website for writing board game AIs

It turns out the Clue thing is just an instance of a more general problem I have. I often get tired of board/card games because it feels tedious playing them. It's like doing random arithmetic problems over and over again. However, I often think about how interesting it would be to write AIs for these games (I've always loved automating boring tasks). For example, although I do actually enjoy playing Pandemic, I think writing an AI for it would be 10x more fun. It's like a classic graph theory problem but with all these constraints thrown in.

Wouldn't it be cool to have a website that facilitated writing AIs for board games and using them to compete against other programmers? It'd be like robot soccer tournaments. For any game, someone could write a server component that would define all the rules and provide an interface for clients. Then anyone could code up an AI client (or a client meant for humans) and plug it in.

This could also be a great tool for teaching programming, another one of my interests. You could gradually introduce someone to programming by giving them most of the components of an AI written with bits left out for them to fill in. As the student progresses you could have them code more and more of the AI.

Project 2: Rewrite Smart Shuffle Player using React Native and Clojurescript

Smart Shuffle Player is what I called an app I built during my undergrad. I took an open-source Android music app and inserted my own recommendation algorithm to make it select which song to play next. It's basically Frankenstein's monster. It was good enough for an undergrad research project, and I've even used it as my main music app for over a year. But there are lots of improvements to be made. It's not in a position where I would recommend it wholeheartedly.

Refactoring the Java code to make it nice would be no bueno—I'd rather rewrite the thing in Clojure. Plus, there's already (relatively) good support for writing cross-platform, native mobile apps with React Native and ClojureScript, so I could make an iOS version too.

The rest of the curriculum

These projects would give me great exposure to the three main areas I want to develop expertise in:

  • Web application architecture (in Clojure)
  • Mobile app development (in Clojure)
  • AI/machine learning/statistics (in Clojure)

As I work on these projects, I have some ideas for supplemental learning activities:

  • Read lots of source code for Clojure web libraries, like Pedestal (HTTP routing), Sente (websockets) and Immutant (web server).

  • Work through Doing Bayesian Data Analysis (a text book I bought a while ago but haven't touched).

  • Finish SICP (the Bible of Lisp. I've read most of it but haven't done any of the exercises).

  • Work on Hacker Rank problems (partly to get practice doing fancy algorithmic things that usually don't come up but also for fun. It's important to take time to enjoy a skill instead of only doing deliberate practice).

I'll also include time for some general ed. I have a never-ending queue of books I'd like to get through, and I love writing.

As time goes on I'll likely modify the curriculum. But even as I've thought about what I have so far, I really think this'll turn into a great experience. I don't even see it as simply "startup prep." Even if I wasn't planning to be a startup founder, this would be a great opportunity to take my skills to the nextLevel++ and launch me into the rest of my career. I have a somewhat mellow complexion, so I don't get excited easily. But I'm excited about this.


Subscribe to be notified about new posts and project updates.