loading ekansh.dev
∇²ψ + (2m/ℏ²)(E−V)ψ = 0
det(A − λI) = 0
P(A|B) = P(B|A)·P(A) / P(B)
∮ E·dA = Q/ε₀
θ̈ + (g/L)sin θ = 0
XOR(p,q) ⊕ hash(k)
lim(n→∞) (1 + 1/n)ⁿ = e
EKANSH DAVE — PORTFOLIO

Ekansh
Dave

Researcher, builder, and competitor obsessed with first-principles reasoning — from materials science at Johns Hopkins to neural nets from scratch. Math is the language everything else is written in.

LIVE SIMULATION RUNNING
DOUBLE PENDULUM
BROWNIAN MOTION
FOURIER
01 · ABOUT

who I am

I'm Ekansh Dave — a student researcher, builder, and competitive debater interested in problems that sit at disciplinary boundaries. I've done materials science research at Johns Hopkins University, built neural networks from mathematical axioms, and argued national-level policy debate.

I think the most interesting problems are the ones where rigorous mathematics meets physical reality — whether that's crystal defect formation, gradient descent convergence, or cryptographic hardness assumptions. The common thread is structure underneath apparent complexity.

When I'm not in the lab or at a keyboard, I'm likely reading about number theory, working through competition math, or preparing debate cases that require the same kind of careful argumentation as a proof.

// DOUBLE PENDULUM — CHAOS FROM SIMPLE RULES
θ̈₁ = [−g(2m₁+m₂)sinθ₁ − m₂g·sin(θ₁−2θ₂)] / ...
L(q,q̇) = T − V = ½m(ẋ²+ẏ²) − mgy
Deterministic equations. Chaotic trajectories. Small changes diverge exponentially.
mathematics & physics
calculus
92
linear algebra
87
probability
83
comp. math
85
CS & engineering
python
91
algorithms
84
cybersecurity
72
C / systems
65
machine learning
pytorch
80
backprop
84
transformers
74
debate
policy
90
LD
82
02 · EXPERIENCE

research & work

full experience →
JOHNS HOPKINS UNIVERSITY
2024 – Present
Baltimore, MD
Undergraduate Research Assistant — Materials Science
Research in the Department of Materials Science and Engineering. Focused on computational modeling of crystalline defect formation and propagation. Used Python-based simulation pipelines to study dislocation dynamics and material failure under stress — applied linear algebra and differential equations directly to physical systems.
materials science python computational modeling linear algebra simulation
INDEPENDENT
2023 – Present
Self-directed
ML Research — Neural Architectures from Scratch
Self-directed research into deep learning fundamentals. Implemented neural networks, optimizers, and the transformer architecture from mathematical first principles — no high-level APIs. Produced detailed mathematical writeups connecting implementations to theory.
deep learning backpropagation pytorch attention
03 · PROJECTS

selected work

all projects →
FEATURED · MACHINE LEARNING
Neural Network from Scratch
Complete deep learning library in pure NumPy. Forward propagation, backpropagation via chain rule, SGD/Adam/RMSProp, batch norm, dropout — everything derived from ∂L/∂w. Built to understand the black box, not just use it.
backpropdeep learningnumpychain rule
01
Double Pendulum Sim
Interactive chaotic system simulator. Lagrangian mechanics, Runge-Kutta 4th order integration, sensitivity to initial conditions visualized as diverging trajectories.
chaos theoryRK4Lagrangian
02
Crypto Toolkit
RSA, AES, and elliptic curve implementations from scratch. Modular arithmetic, extended Euclidean algorithm, discrete log problem — math-first cryptography.
cryptographynumber theorypython
∑ λᵢxᵢ
03
Transformer from Scratch
Multi-head attention, positional encoding, FFN layers, layer normalization — built from Vaswani et al. math. No Hugging Face, no shortcuts.
attentionpytorchNLP
04 · CONTACT

reach me

Open to conversations about research, ML, cryptography, math problems I haven't seen, or debate. Response time: 24–48h.

TWITTER / X@ekanshdave
contact.py
# reach out if you're into:
interests = [
"research collaborations",
"ml + math discussion",
"cryptography / ctf",
"hard problems"
]
 
$ python contact.py --send
→ message delivered