With the hype surrounding data science, artificial intelligence (AI), and machine learning (ML), I meet a lot of people who would like to get involved in the field. These individuals often voice their frustration; the journey into this field appears to be gated by extensive prerequisites and long-winded, traditional academic pathways. However, my own path into ML, which spans just over a decade, tells a different story — a narrative not of shortcuts, but of persistence, curiosity, and continuous learning. I want to share my story, in the hopes that it may help or inspire others.
Eleven years ago, I was nearing the end of my business degree; today, I have the privilege of developing innovative AI products at Google. This transition wasn’t quick or easy, but it was possible. Here’s how.
At the end of 2012, while pursuing a business degree at a local college, my encounter with Coursera — a platform offering Massive Open Online Courses (MOOCs) — marked the beginning of my AI journey. Initially dabbling in various subjects, I found myself drawn to the more technical courses, such as statistics and computer science. Roger Peng’s ‘Computing for Data Analysis’ course introduced me to R, a statistical programming language, igniting a passion for both R and data analysis. This passion was further fueled by Jeff Leek’s ‘Data Analysis’ course, where I first encountered ML and concepts like decision trees, bagging, and random forests. Early in 2013, I fit my first ML model.
In early 2013, eager to deepen my understanding, I began to incorporate R into my college assignments, explore Kaggle competitions, and tackle side projects. One particular Kaggle competition — predicting salaries from job listings — proved particularly challenging due to my lack of experience in natural language processing (NLP) — the branch of AI that focuses on enabling computers to understand, interpret, and generate human language. Despite spending several evenings playing with the dataset, I did not make much progress. In fact, I struggled with manipulating the dataset, as it was quite large for that time. I never made a submission to Kaggle for that competition. This experience felt akin to venturing in the wrong direction at the start of an RPG and getting wiped out by a high-level dragon. It left me thinking, ‘I’ll return later and slay this dragon.’
Around the same time, I stumbled upon ‘The Elements of Statistical Learning’, a foundational textbook in the field of statistical learning. Although I managed to get through the book, I understood only a fraction of it.
I continued to expand my skills by enrolling in Coursera’s ‘Introduction to Data Science’ course by Bill Howe. Unfortunately, I flunked the first assignment — a sentiment analysis task — due to my complete lack of Python and associated programming skills.
In September 2013, leveraging my internship experience, I was offered a full-time role as a business analyst. I had originally planned to return to university to pursue an MSc, but I decided to enter the workforce immediately, thinking I could always pursue an MSc part-time later.
Despite my rapidly expanding skill set, my non-traditional academic background still presented obstacles. I faced rejections for internal transfers from some potential hiring managers in technical teams, attributed to my lack of a formal quantitative background. This issue also complicated my attempts to gain admission to formal academic programs. After being denied admission to an MSc in Business Analytics program because of my ‘non-quantitative undergraduate degree’, I pursued a Postgraduate Certificate in Statistics to help bridge the educational gap.
When the subsequent iteration of ‘Introduction to Data Science’ on Coursera was offered in September 2014, I blasted right through it, completing the assignment that had defeated me a year before in just a matter of hours.
My perseverance in formal education also paid off. In 2015, I was accepted into the MSc in Business Analytics program from which I had been previously rejected. For my dissertation in 2017, I revisited the Kaggle dataset that had previously stumped me, this time applying deep learning and NLP techniques. This project, a collaboration with a classmate, incorporated a range of NLP methods, and our final model achieved solid results, placing approximately 5th in the original Kaggle leaderboard. This accomplishment showcased how far I had come.
From 2018 to 2022, my professional focus shifted more toward software engineering, moving away from data science and ML. Despite this, data science and ML remained my North Star, guiding me to develop the strong programming skills necessary to fully realize my ML and AI aspirations.
In 2022, after several years of working on data processing infrastructure, I began searching for my next role, aiming to return to ML. I secured the perfect position as an ML engineer, working on a new Google Cloud AI product.
In February 2023, I embarked on another program of part-time study, this time pursuing a diploma in Data Science. The first semester included a module on Computational Statistics, for which ‘The Elements of Statistical Learning’ was the prescribed textbook. As I prepared for the semester’s exam, I revisited this book, finding that I understood almost all of it — a significant improvement from my first attempt ten years earlier.
Inspired by the release of OpenAI’s updated embeddings in January 2024, I revisited the same old Kaggle job salary dataset, eager to leverage these advancements. I developed a single deep learning model that potentially could have placed second. By integrating the predictions of this model with those from two others — each trained under slightly varied conditions — I created an ensemble that would have secured a hypothetical first-place finish on the Kaggle leaderboard. This achievement marked a significant milestone in my personal journey.
My journey from a non-technical background to a career in ML was neither quick nor straightforward. It demanded patience, a readiness to learn from the basics, and an unwavering commitment to continuous learning. To those aspiring to enter the field, remember: the path may not be easy, but it is certainly achievable. Embrace challenges as opportunities for growth, accept that some problems are not yet within your reach (make a tactical retreat and come back later) and never underestimate the importance of curiosity and persistence.
In the realm of AI and ML, every setback presents a learning opportunity, and every challenge is a step toward mastery. I hope my story serves as a testament to the power of perseverance and the endless possibilities that await those willing to embark on their own journey of discovery.