Sleep ∉ Book Writing

My beautiful 9PM-7AM work schedule

What?

I expected to get a lot of things out of my Boeing Internship: a blast of the corporate office life, long Zoom meetings in code review with my mentor, and a growing exhaustion with the nearly two-hour commute to the site. All of these things happened, but I also ended up writing a book with my mentor.

Pretty unexpected. For context, my mentor loves learning, and had recently finished up a masters program at CMU (one of nearly ten of his graduate degrees) involving time series forecasting, or the practice of modeling and making predictions with data over time. Stock prices, temperatures, energy usage, that kind of thing. So, when he ran into me, an intern with a plethora experiences working with Python and data-sciency ML, he asked me if I wanted to help him turn the daunting topic of time series forecasting into an accessible textbook that anyone with a high-school background could make sense of, I jumped at the opportunity. I mean, who gets the chance to write a book in high school?

And the monkey's paw curled.

I can relate to Icarus

The initial stages of the project went well. I wrote outlines, came up with a title, and started modifying an existing codebase my mentor had made to make extra diagrams and examples for readers to understand. I was going to take them on a crash course throughout modeling, whether it be older statistical models like ARIMA, simpler ML implementations like linear regression and random forests, to more modern models like gradient boosting. Time series forecasting was a great way to deliver fundamental ML concepts, like good data selection/cleaning, feature selection, and assessing models with overfitting in mind in an easily understandable way. In my own previous experiences, I felt like it was easy for beginners to make their first "Cats VS Dogs" image classification model and be completely abstracted away from the fundamentals. I can't really blame them, as while image models are interesting, understanding these very dense neural networks, in incomprehensibly high dimensions, with convolutions, batchnorm, activation functions galore can be intimidating, especially for people just learning how to code in Python.

The intentions were there, but as I started writing the first few chapters I realized something. It was just too much work. We had a deadline of 8 weeks, so we could finish and present to the rest of the Boeing internship program, and maybe in isolation it could have been straightforward. But I had so many other things taking up my time, like taking Physics online, zoom meetings with my research mentor (see: 8 H100s and a Dream), volunteering to develop websites for artists in Tanzania, Africa, all in all I was swamped. After a while, my productivity hit a wall and I felt pretty discouraged. For each chapter, I had to learn about the topics, write example code, make diagrams, and consult with my mentor about the validity of what I was writing.

Seasonal Decomposition? More like sanity decomposition

Stockholm Syndrome?

"Just Work Harder" is not a mantra I would recommend to a friend, but in this case I really just needed to work harder. Thankfully, something about tackling a challenge you thought was too much for yourself, giving everything your undivided attention for a week or so is incredibly fun. Through a series of all-nighters, using my laptop on the commute to work, and other ways to squeeze every bit of productivity I had left in me, I blitzed through the last few chapters, which involved the meat of a lot of the machine learning methods. The next step was to export everything to a Word file (whose formatting was honestly more daunting than writing the book itself). Lastly, I hopped onto my favorite open-source "not Photoshop" software Gimp, and put together a cover page... which didn't meet Amazon's sizing requirements. After a few more iterations, and figuring out layers so I could overlay the required size over the new title, I finally finished the book. I breathed a sigh of relief as I hopped on a Webex call with my mentor who was equal parts a bit surprised at my sudden burst of progress and proud of my results.

We Get Published

Eventually, I got everything uploaded to Amazon, and after a (few) hiccups involving print sizing and Amazon's book legibility requirements, I received the first prints of my book. I think that was the most excited I've been to open one of those familiar Amazon packages. I wasn't buying something from someone else (other than the production of the paper book itself), I was getting a hold of something that I made myself. A few days later, I got to show off my hard-copies to fellow interns and curious employees alike at our intern project poster session, and it was nice getting to reap the benefits of those all-nighters.

While putting my life on the line for this book did lead to me accumulating a few overdue assignments and pushing back my other responsibilities for a bit, I think getting the clarity of going all-in on one project was definitely worth it. Though the sleep-deprived fever dreams I had after my work sessions were not fun...

Whatever! Onto the next big thing.