Data Science or Bust
Two Roads and the woods, maybe?
A journey of a thousand miles starts with one step… or something inspirational like that. I am not much on the inspirational pep talks or the long arduous journey metaphors that make you out to be Frodo and Sam on the way to Mount Doom to destroy the One Ring — when it comes to these sort of posts. I am a lot more of a post-modernist in that regard. With that said, what was my journey and how did I end up on the data science hype train, and why?
First off was the formation of my love of computers that started in the early noughties when my parents bought an old, and supremely silly cow-spotted Gateway PC. I’ve always lived in a very rural area so all we had was dial-up, but I loved exploring the burgeoning consumer internet and learning all that I could. These were the days of Yahoo and Ask Jeeves before Google rose to prominence and everyone still checked email through a slow client. Eventually, in my pre-teenage years, my parents would buy me my own Dell tower and one of the old CRT monitors. I found gaming and fell in love with the fantasy worlds that video games led me to explore, but as someone with a naturally curious and creative bent, I found myself picking at the edges of video games. The urge to create pushed me to find out what made a game tick.
As I got older I started finding the inner-workings of a PC more and more interesting, and would soon catch the bug big time when my parents purchased my first GPU (Graphics Processing Unit AKA Video card). I was hooked now on both hardware and software, and curiosity made me dig deeper. I found BASIC, one of the early programming languages, during one of my early marathon internet sessions (you tied up the phone line on dial-up so you had to get all of your interneting in one big go). I dove into several of the variations of BASIC and really started to pull things together, and how far things have come since the english readable, but basic (no pun intended) syntax and structure of BASIC. GOTOs, BREAKs, JUMPs, all sorts of oddities that we luckily don’t have to deal with today populated these ancient languages — but I ate them up.
I outgrew BASIC and started exploring Visual BASIC in high school, but eventually, that would change to the more sophisticated (I thought at least) C++. Through highschool I stuck to learning some of the basic building blocks of Object-Oriented Programming and teaching myself the syntax and uses of C++. On top of this, I was able to take 4 semesters worth of CISCO Networking which granted me so much knowledge on the server and administration side of things, as well as an advanced Multimedia Design class which solidified my knowledge of front ends and customer-facing services through the web.
After high school I attended college with plans to study, not computer science as you may have guessed by now, but legal studies. I spent about 2 years working towards a paralegal certification and eventually landed a brief gig at a law firm doing clerical work for an attorney specializing in coal-related Black Lung cases. I learned quickly that a law office was not the place for me, and while I adored the process of legal writing and research I hated the atmosphere and the lack of real agency. This really drove me to want to do something else, so I moved and started college and work in a small college town. I worked through a few semesters of course work on computer science while working full-time. Circumstances, unfortunately, led me to leave college and only work.
I’d been working in restaurant leadership for a while and COVID hit. Taking a voluntary leave for my health allowed me a lot of free time and the ability to better myself. I chose to work through a 5 month, intensive Bootcamp was my best bet to make this time off work for me. This allowed me to take my love of computers and computer science and combine it with my experience in data-driven decision making and optimization by applying it to Data Science. Now, 5 months on and on the verge of graduation I have come a long way. I have achieved so much and I am very proud of what I have learned and the work I have done.
My two greatest accomplishments are two in-depth deep learning projects. These two projects allowed me to push my abilities by going above and beyond the curriculum and teaching myself PyTorch and numerous other libraries ancillary to deep learning.
The first project is a computer vision project that using a convolutional neural network to detect pneumonia in the lungs of small children (ages 5 and under). This taught me the basics of PyTorch and the complications of working with image data, which allowed me to stretch my OOP legs and implement the network as a class and fairly self-contained. Pneumonia Computer Vision
My crowning achievement is my AI-Generated review classifier. Built again with PyTorch it allowed me to get a feel for Natural language and the specific processes that are required for text data. In order to build this classifier, I utilized a dataset of 51 million amazon book reviews of which I sampled 50,000 reviews that trained a generator based on the GPT2 architecture for transfer learning. I then used the sampled reviews and the generated reviews and was able to train the classifier to classify AI-generated reviews at an approximate 80% accuracy. AI Generated Review Classifier
Now we are here! I am excited to apply the skills that I have developed and work my way towards something greater.
Originally published at http://github.com.
