My name is Vinjai Vale. I am pursuing a math major and CS masters at Stanford (expected graduation June 2022). You can contact me at [firstname]@stanford.edu.
My research interests include deep generative models, value-aligned development/application of artificial intelligence, and existential risks. I am a Co-President of the Stanford chapter of the Association for Computing Machinery, and co-founded the Stanford Existential Risks Initiative. I enjoy ultimate frisbee, rock-climbing, and StarCraft II.
Recent work
-
Scale (2021): ML infra + full-stack software engineering. Conceptualized, researched, and built end-to-end the new active learning feature in the rapidly growing Nucleus product, which helps Nucleus customers curate unlabeled data to prioritize the highest-value data to send to labeling.
-
Skiff (2020): full-stack software engineering. Worked on key features like content expiration backend and e2e group workspace UI.
-
Jump Trading (2020): quantitative research. Won the interns' algorithmic trading strategies competition and rotated on two trading teams.
-
AWS Research (Deep Learning + Digital Signal Processing Group; 2019): ML research. Developed deep latent variable models for end-to-end speech audio enhancement. Published in ICASSP 2021 as co-first author.
-
Ng lab (2019): ML research. understanding and predicting treatment response in depression patients. Published in JAMA Network Open.
-
Helm.ai (2018): ML research. Developed unsupervised learning algorithms for camera-based depth perception in autonomous vehicles.
Interesting courses
Math 116: Complex Analysis
Math 136: Stochastic Processes
Math 154: Algebraic Number Theory
Math 159: Discrete Probabilistic Methods
Math 172: Lebesgue Integration and Fourier Analysis
Math 228: Stochastic Methods in Engineering
CS 107, 110: Computer Systems
CS 166: Data Structures
CS 168: The Modern Algorithmic Toolbox
CS 224N: Natural Language Processing
CS 228: Probabilistic Graphical Models
CS 234: Reinforcement Learning
CS 236: Deep Generative Models
CS 255: Introduction to Cryptography
CS 261: Optimization and Algorithmic Paradigms
Law 4039: Regulating Artificial Intelligence
Projects
-
CS 236 final project: introduced methods to learn multiple orderings simultaneously in autoregressive models, and demonstrated that this leads to better inpainting results and sub-linear inference time. Read our paper.
-
CS 166 final project: explored dynamic optimality and analyzed the tango tree data structure in theory and practice. Read our paper.
-
Won the MIT Battlecode competition in 2019, won second place in 2020, and was a finalist in 2021 with team smite. See our 2020 repository, 2019 code and 2019 strategy report.
High school
Graduated from Phillips Exeter Academy in 2018, where I was recognized with the Faculty Prize for Academic Excellence for holding the first rank.
Competed in the 2017 Regeneron Science Talent Search, where I finished in 7th place. Read my paper.
Founded the Phillips Exeter Academy Puzzle Hunt and led the effort to organize three annual 48-hour hunts. See our puzzles and this article. The hunt has become a school tradition and is now on its sixth year.
Many years ago I was active in math competitions. I won first place in the 2015 USA Junior Math Olympiad and went to the Math Olympiad Summer Program. I also wrote a book, 111 Problems in Algebra and Number Theory, for aspiring Olympiad contestants.