I submitted my application to Stanford’s Knight-Hennessy Scholars program today. The application was truly a group effort - if you helped me and are reading this, thank you so much - and I had so much fun constructing my answers to the prompts, I wanted to share what I finally submitted.
My intentions after earning my doctorate will depend heavily on how the two activities I’m currently pursuing, research and my startup, progress. If my doctoral research in artificial intelligence is outstanding, I want to continue a research career, with the immediate next step of applying to postdoc positions or perhaps directly to tenure-track faculty positions. If my startup, which aims to empower young Americans to build home equity via fractional mortgages on shared properties, succeeds, I want to continue leading it. If the startup fails, I want to begin working on another startup, ideally one that commercializes my research in some capacity. Long term, I want to span academia and entrepreneurship, working in highly collaborative environments with bright minds motivated by a shared mission to develop and apply novel solutions to significant open challenges.
Working at Uber taught me three lessons. First, my intensity is proportional to the magnitude of the problem. Second, working with motivated teammates amplifies my intensity; I thrive on collaborating for a shared vision. Third, my ability to make a difference grows with the significance of the problem. I know the problems in my field, but my impact will be greater if I can find bigger problems, which necessitates looking beyond my domain. I delight in learning about complex problems, brainstorming with others how to attack them and then working together relentlessly to solve them.
The Knight-Hennessy Scholars program provides an unparalleled opportunity to excel in my field while building connections to the world. The program forges friendships across disciplines and backgrounds; this highly collaborative approach would place me alongside passionate leaders in their own fields, to bring out the best in me.
Before finals of Congressional Debate’s Tournament of Champions, I didn’t have time to practice my speech against America’s war in Afghanistan. Unfortunately, that topic was drawn. My dad/coach braced for disaster. Speaking extemporaneously, I surgically refuted all preceding speakers, ending with a punchline that left the audience howling and placing third as a junior.
My only opportunity to publish before PhD applications was the premier machine learning conference. I worked fourteen hour days, incurring elbow pain so lancing it rendered one hand useless. We discovered a flaw in my proof. Against the clock, I fixed the flaw, and the paper was accepted!
Towards the end of my term as CFO/COO of UC Davis’s student government, I had grown frustrated with political interference in my businesses. After losing a key battle, I yielded to rage and impulsivity and resigned in protest. I abandoned my standards, boss and business managers.
Winning the national championship was the start of my descent. I had spent every weekend in high school honing my ability to speak and debate, only to realize at graduation that this activity - my joy, my community, my identity - was permanently over.
At college, lecture and problem sets couldn’t compare to the intensity of consuming knowledge, analyzing and arguing its merits. Adrift, I switched from biomedical engineering to economics to consumer science to computer science and statistics. I explored business by managing UC Davis’s student government’s 26 businesses and $12 million budget, but grew disheartened with political interference in my businesses’ operations and resigned in protest. My coursework suffered, but all I wanted was a project worth championing. I discovered that other universities permit undergraduates to teach their own courses; wanting that opportunity, I spent two years lobbying UC Davis’s Academic Senate to approve a similar program, and then taught three courses and guided seven undergraduates through teaching their own. I graduated with the Computer Science Department’s Outstanding Graduating Senior award, but I was still lost and now depressed.
For eleven months, I applied to jobs without success, a crushing experience. Research interested me, but I had no graduate degree and no hope of obtaining one with a transcript peppered with Fs. I started reading publications in deep learning and reinforcement learning, implementing the algorithms and writing an explanatory blog that surpassed twenty-thousand unique readers in a month. I joined Stanford’s Computational Neuroscience Journal Club and presented a paper. With self-taught knowledge, I secured a summer internship in which I proposed and implemented a deep learning-based DNA sequencing algorithm, outperforming the gold-standard and earning a patent.
My journal club friends encouraged me to continue learning, so I enrolled at University College London to study cognitive neuroscience. I learned a tremendous amount about human neuroimaging and the field’s research questions. I took advantage of UCL’s proximity to DeepMind by creating UCL’s Artificial Intelligence Journal Club, inviting researchers to present their research and hosting discussions afterwards at the pub on promising directions. I loved every minute of those intimate discussions.
By the end of my program, I was itching to learn how machine learning research was conducted and applied at scale. A journal club speaker referred me to Uber and I accepted Uber’s offer. Uber was a terrific learning experience, but the problems we solved weren’t fundamental and my impact as an individual contributor was marginal. I realized that by staying at Uber, I was well positioned for a comfortable career, but I wanted to love what I did and not love leaving the office.
I returned to academia, this time in theoretical neuroscience and machine learning. In my graduate courses at Harvard and MIT, I eagerly devoured the material. Within 5 months, I published a paper at a top tier machine learning conference, an exceptional pace. I have two more papers on the way, and I’m aiming for five total by the end of my two year program. Facing a lower income trajectory due to attending graduate school, a friend and I crafted a business model to enable young professionals to convert rent into home equity, and we enrolled in Harvard’s startup incubator to build our company. Our goal is to make housing accessible and affordable again for young Americans.
Since moving to Cambridge, I’ve been working 11 hour days, weekends included, for less than minimum wage because I love my research, startup, education, professors and peers. To Jobs, I can’t claim to have found what I love to do, but I haven’t settled either and I’m getting closer every year.
I’m interested in developing mathematical and computational models to explain how intelligence can arise from biological and artificial neural networks. My research aims to answer diverse questions in theoretical neuroscience such as why and how does distributional reinforcement learning confer an advantage, what computational principles guide the formation and function of memory engrams, and how to build interpretable models of high dimensional, non-linear systems with many dynamical degrees of freedom. To accomplish this, I closely examine experimental findings to drive interesting theoretical insights and leverage tools from Bayesian inference, reinforcement learning, machine learning, dynamical systems and control theory.
After finishing my Bachelors’ degrees, I wanted a job as a research scientist or research engineer, but I lacked the requisite skills. Over the course of eleven months, I taught myself deep learning and reinforcement learning and started an online blog authoring detailed walkthroughs of recent publications in deep learning and computational neuroscience; within a month, my explanations surpassed twenty thousand readers and topped HackerNews. I also joined Stanford’s Computational Neuroscience Journal Club, where I presented one of my posts. I also honed my programming skills by completing two personal machine learning projects, one replicating a paper and another applying transfer learning in computer vision.tags: applications - Stanford - 2020