Dehn invariant zero tetrahedra
A comprehensive list of currently-known tetrahedral families with Dehn invariant zero. This is the result of a UROP conducted with Prof. Bjorn Poonen and Kimi Sun.
About Me!
I'm an MEng student at MIT LIDS, advised by Prof. Marzyeh Ghassemi.
I previously completed my undergrad in 2024 from MIT as well, where I doubled in Math (18) and AI (6-4).
Very broadly speaking, I am passionate about the intersection of theory and computation. My thesis centers on developing robust deep learning models - using theoretical insights to address and mitigate the impact of noisy data, with applications in healthcare diagnostics. I work to establish theoretical guarantees and empirical probing methods for robustness to noisy labels.
I am also interested in applying algorithmic and machine learning tools to uncover patterns in number theory. This includes converting theoretical results to efficient algorithms, and thinking about which architectures capture the correct "symmetry" in mathematical data (e.g., equivariant learning).
I am involved with Algeria's national mathematics olympiads, and I also previously served as problem staff at HMMT.
Here's my CV if you would like to have a look at it. A list of classes can be found here, while a list of projects can be found below.
Resources
The following are some resources I've written up over time.
Adversarially-Resistant Bloom-Filters
In this group project, we analyzed the memory usage (in bits) needed to maintain an adversarially-resistant bloom filter, where resilience is defined in terms of a security game. We improved known lower bounds to get near-optimal memory usage in sublinear cases.
Discrete Logarithm Problem
Notes on the discrete logarithm problem, particularly motivated by the Diffie-Hellman exchange protocol in cryptography. This was written as a final paper for 18.704 - Seminar in Algebra.
Dirichlet Convolution and Möbius Inversion
Notes on arithmetic functions and Dirichlet convolutions, particularly aimed towards motivating the latter as well as demysifying the Möbius inversion. I wrote this as undergraduate assistant for the Spring 2022 offering of 18.781 (MIT's elementary number theory class), taught by Prof. Ju-Lee Kim. I wrote these because I found that most textbooks' treatments of the subject to be relatively unmotivated and/or confusing. I had struggled for years with fully grasping what Möbius inversion was really about, and so I hoped that this would help students avoid the issue.
Prime Number Theorem
Notes from 18.112 on the modern proof of the prime number theorem, along with some elementary consequences of the theorem.
Crash Course in Topology
Elementary point-set topology notes in Andrew Lin's LaTeX package.
18.022 Notes
This is frankly just a prone-to-error summary of important formulae and definitions that I spiced up by including as much analysis as I could.
Projects
The following are some practical projects I've done over time, loosely organized by topic, along with a short description. Research projects are listed on my CV.
Signal Processing
- JPEG Encoder
Implemented an efficient JPEG encoder which inputs an image in raw form as brightness values and outputs a JPEG image. - Separating Harmonies
Used frequency-domain methods to design a Python program which separates bass, melody, and harmony components of a 3-part harmony. - Echo Cancellation
Employed convolutional methods in the time domain to de-echo a real-life audio message. The same method was applied but in the frequency-domain for a more efficient implementation. - Note Detection in Chords
Converted a signal to the frequency-domain representation using Fourier transform, which was then analyzed to identify the corresponding notes in the original signal.
Programming
- Solver for CNF Formulae Implemented an efficient solver for Boolean satisfiability problems.
- Symbolic Algebra
Used Object-Oriented Programming (Python) to build a symbolic algebra system. - N-dimensional Minesweeper Wrote and improved Python code which simulates a playable N-dimensional Minesweeper game (including the rules, moves, and the status of the board).
Cryptography and Security
- CRIME attack against gzip compression
Conducted CRIME attack which infers the contents of a message by observing the change in size of the compressed payload. This attack specifically targeted the gzip compression algorithm. - Broadcast attack against RSA
Implemented the broadcast attack against the RSA: this exploits encrypting the same message under different public keys. - One-time pad encryption with same key
Theory teaches us that it is insecure to use the same key twice for symmetric encryption. This cryptanalytic attack demonstrates this: we use a library of known English words to completely retrieve two original message that were encrypted using the same key under symmetric encryption. - Memory Usage of Adversarially-Resistant Bloom Filters
In this group project, we analyzed the memory usage (in bits) needed to maintain an adversarially-resistant bloom filter, where resilience is defined in terms of a security game. We improved known lower bounds to get near-optimal memory usage in sublinear cases.
Other
- Imaginary Quadratic Orders
Implemented an efficient algorithm to enumerate all imaginary quadratic orders of class number less than a given bound, assuming the Generalized Riemann Hypothesis. - ML Classification of Elliptic Curves
Trained machine learning models to predict the Sato-Tate group of elliptic curves using Scikit Learn.
Some (not so professional) photos I've taken over tyears
