Long Bio

2012-17

*Self-Study, Research Projects, and Data Science*

Growing up in South Bend, Indiana, I always enjoyed math. I initially focused on sports, but once I realized that I had more talent and fun with math, I leaned into it by self-studying MIT's undergraduate math/physics curricula. This started during the summer of 2012, after my sophomore year in high school. I had taken precalculus at school that year and encountered a bit of calculus in the spring, and it seemed really interesting so I figured I'd try teaching myself the rest over the summer using various online resources. Self-study turned out to be way more efficient than I was used to at school, and it was incredibly fun making progress so quickly. Once I got to optimization and related rates, I was completely hooked, and voluntarily holed up in my room working out math problems.

After calculus, I moved on to Linear Algebra and Multivariable Calculus through MIT OpenCourseWare, and once school started up again in the fall, I just kept on going with the rest of undergraduate math (plus half of undergraduate physics and a bit of mathy coding). During my junior and senior years of high school I also worked at Mathnasium on evenings and weekends and did some science fair projects in experimental physics.

In 2014 I received a full ride academic scholarship (Lilly Scholarship) to the University of Notre Dame, majoring in mathematics, and became obsessed with modeling the human brain as a weighted directed graph. This led to some fun, but not particularly fruitful, research projects. During my first year I worked on a bottom-up approach that became intractable beyond simple cases, and I spent the following summer in Los Alamos working on a top-down approach attempting to produce biologically realistic brain waves using spiking neurons in a deep neural network. I got the sense that brain modeling was not a great problem to devote my life to.

During the spring of my second year (2016) I interned as a data scientist at Aunalytics while taking a variety of grad classes across pure and applied math. I lost interest in academia, and Aunalytics was happy for me to work full-time throughout the remainder of my degree, so that's what I did. I built models and pipelines for churn prediction, conducted user segmentation studies in financial services / digital news / general subscription services, and investigated whether tools from topological data analysis could be useful in any analytics pipelines.

At some point I realized that quantitative modeling was most valuable when paired with domain expertise, and although I was grateful to have broken into data science so early, I wasn't excited by the conventional problem domains. I was only excited by tutoring and music production, and it was unclear how to leverage mathematical modeling in either domain. But it was clear that tutoring could keep me afloat if I moved to a bigger city. Los Angeles was the obvious choice since it's where my girlfriend (now fiancé) Sanjana was studying.

2018-23

*Math Academy and Eurisko*

After graduating from Notre Dame at the end of 2017 I moved to Los Angeles and took on scattered work in math education: tutoring, teaching high school and test prep, developing content for math websites, and even writing several textbooks for fun. I also picked up a master's in computer science from Georgia Tech (it was only $7k, so, duh!). Gradually, all my work converged at the same organization, Math Academy, a highly accelerated 6-12th grade math program where 8th graders took AP Calculus BC and high schoolers studied a full undergraduate math curriculum. Math Academy was founded by Jason and Sandy Roberts and ran on top of educational software built by Jason (and fully self-funded out of Jason and Sandy's own pocket, completely separate from the school program).

I got involved at the core of Math Academy's software during the summer of 2019. At that time, the software had existed for several years as a tool that Math Academy instructors used to create and grade assignments -- they would manually select problems from the database (which contained a mountain of content written by Math Academy's team of PhD mathematicians), students would complete the problems online for homework, and the software would automatically grade the assignments and keep track of each student's grades. But it was very time-consuming to manually choose a mix of problems that covered the multiple topics taught during class each day and also provided spaced review on previously-learned topics.

It was clear that while students were learning an incredible amount of math, giving the same assignment to each student in a class still left a lot of learning efficiency unrealized: even within a single class, different students had different strengths and weaknesses, had different sets of topics that they were ready to learn, and needed different amounts of practice on different topics to reach a sufficient level of mastery. Giving each student the same assignment virtually guaranteed that every student wasted lots of time being bored or lost -- either way, not actually learning. To solve this problem, different students would need to learn different topics at different times, and get different amounts of practice (including review) on those topics, and each student's learning plan would need to continually adapt to their individual performance.

The need for fully-individualized learning, as well as other needs (e.g., financial sustainability and quality control) led Jason and Sandy to the realization that the only way forward was for the system to become a fully-automated standalone online learning platform, commercially available to the general public. Aware of my background, Jason asked me to develop an algorithm that would automatically assign personalized learning tasks (personalized to each student's individual knowledge profile) while leveraging effective learning techniques like mastery learning, spaced repetition, and interleaving. By the end of summer the project was a success, upgrading the software from a manual assignment creation tool to a fully-automated adaptive learning system that could effectively support independent learners (without any teacher).

We kept iterating on it, year after year:

- During the 2019-20 school year, we started out with a single independent learner, a student who was previously in Math Academy and had moved to another state. She learned AP Calculus BC using only the system (i.e. no external help) and got a 5 on the AP exam.
- At the start of the COVID-19 pandemic in spring 2020, I moved in with Jason and his family to quarantine, which led to a makeshift startup incubator experience run out of his living room. We worked every waking hour, well into the night, every day including weekends, so that by the end of summer we were ready to run entire school classes on the automated system.
- During the 2020-21 school year and COVID-19 pandemic, the automated system proved to be significantly more effective than traditional remote instruction, and by spring 2021 nearly all of Math Academy's school classes were running on it (a couple of which I personally managed until 2023).
- During the 2021-22 school year, even after school was back in person, we reached the point that the system was 4x as efficient as traditional in-person classes covering the same material. Seemingly impossible things started happening like some highly motivated 6th graders (who started midway through Prealgebra) completing all of what is typically high school math (Algebra I, Geometry, Algebra II, Precalculus) and starting AP Calculus BC within a single school year. Math Academy's AP Calculus BC exam scores scores rose, with most students passing the exam and most students who passed receiving the maximum score possible (5 out of 5). Four other students took AP Calculus BC on our system, completely independent of a classroom, and all but one of them scored a 5 on the AP exam (the other one received a 4).
- During the 2022-23 school year, we opened up mathacademy.com commercially to the world at large, became WASC accredited, grew to hundreds of commercial users, and (nearly) hit operating break-even.

Over these years I simultaneously did a lot of work with Math Academy's school program, including developing what was, during its operation from 2020-23, the most advanced high school math/CS sequence in the USA. In these courses, I scaffolded high school students up to doing masters/PhD-level coursework (reproducing academic research papers in artificial intelligence, building everything from scratch in Python). We called the program

*Eurisko*, which is Greek for "I discover," and is the namesake of an AI system from the 1980s that won a particular game competition twice in a row, even when the rules were changed in an attempt to handicap it.

2023-present

*Scaling Math Academy*

During the summer of 2023 I left Math Academy's school program and relocated to Boston, where I continue to work on our product in full force.