nitmb-poster.jpgNSF–Simons NITMB MathBio Convergence Conference
I presented graph-theoretic analyses of aperiodic polykite tessellations and how they inform models of viral capsid assembly, with competitive travel funding.
PhD student, Applied Mathematics · Case Western Reserve University
How do simple geometric rules build a virus, how does chance decide whether an epidemic takes off, and how can data help us predict what comes next?
Research interests
I use mathematics and data to understand complex systems, from the tilings that build viral shells and the randomness that decides whether an outbreak survives, to neural networks that forecast energy demand.
I am a PhD student in Applied Mathematics at Case Western Reserve University. My work combines graph theory and aperiodic tilings, stochastic modeling, and machine learning. I am interested in how viruses assemble, how diseases spread, and how models learned from data can make reliable predictions. My earlier work used recurrent neural networks to forecast electric energy consumption.
Before CWRU, I earned an MS in Applied Mathematics at the University of Delaware, where I worked with Prof. John R. Jungck on quasicrystalline models of viral capsids. Before that, I earned a BS and MS in Mathematics at Jahangirnagar University in Bangladesh.
nitmb-poster.jpgI presented graph-theoretic analyses of aperiodic polykite tessellations and how they inform models of viral capsid assembly, with competitive travel funding.


The NSF Internship Network in the Mathematical Sciences trained me in Python, statistics, and machine learning for work beyond academia. I worked on team projects with mentors Prof. Fadil Santosa and Dr. Martin-Daniel Lacasse.
3mt.jpgMy research in three minutes and one slide. Watch the talk
capsid-models.jpgOur junction analysis guided where magnets go on 3D-printed capsomere models, built with Prof. Donald Plante (UNH).
hpc-seminar.jpgAs chapter treasurer, I organized a seminar introducing graduate students to HPC at CWRU.
outreach.jpgI co-led a seven-week workshop on knot theory and DNA, fractals, Voronoi diagrams, and epidemic models for students from underrepresented groups.
gbd.jpgI review GBD 2023 manuscripts on disease burden worldwide, from brain cancer to respiratory infections. I completed IHME's Introduction to GBD course. Certificate
Many viruses build their shells from repeated protein tiles. With Prof. Jungck, I model quasicrystalline capsids using the newly discovered aperiodic polykite "einstein" tiles: hats, turtles, hares, and squirrels.

Even when \(R_0 > 1\), an outbreak can die out by chance. I compare early extinction in direct-transmission (SEIR) and vector-borne (Ross–Macdonald) models at matched \(R_0\), using Gillespie simulation and branching-process theory.
epidemic.pngHow much electricity will a home, a city, or a country need tomorrow? With collaborators, I used recurrent neural networks to learn patterns in past energy use and predict future consumption, work that supports planning for smart homes and power grids.
ml.png
poster-thumb.jpg
3mt-thumb.jpg
Advisor: Prof. John R. Jungck. Passed PhD preliminary exams in Real Analysis, Linear Algebra, and Numerical Analysis.
Mentors: Prof. Fadil Santosa and Dr. Martin-Daniel Lacasse
Quasicrystalline models of viral capsids with aperiodic polykite tiles


Thesis: Poiseuille flow of incompressible fluid, analytical and finite-difference approaches.
With Dr. Kabir Uddin Sikder

Project: machine-learning prediction of solar energy consumption. University Merit Scholarship.

Leadership, competitions, and activities outside the lab. Tap a photo to see it larger.
Why 12 pentagons?
A closed shell is tiled by hexagons and pentagons, with exactly three tiles meeting at every vertex. Show that it must contain exactly 12 pentagons, no matter how many hexagons there are.
Let \(p\) and \(h\) be the numbers of pentagons and hexagons, so \(F = p + h\). Each edge borders two faces and each vertex touches three, so
$$E = \tfrac{1}{2}(5p + 6h), \qquad V = \tfrac{1}{3}(5p + 6h).$$Euler's formula \(V - E + F = 2\) gives
$$\tfrac{1}{3}(5p+6h) - \tfrac{1}{2}(5p+6h) + p + h = 2 \;\Longrightarrow\; \tfrac{p}{6} = 2 \;\Longrightarrow\; p = 12.$$The hexagons cancel completely, so \(h\) can be anything.
Can a growing outbreak still vanish?
An outbreak starts with one infected person. Each infected person, independently, infects 2 new people with probability \(p = \tfrac34\), and nobody with probability \(\tfrac14\). On average each case causes \(R_0 = 2p = 1.5\) new cases, so the outbreak "should" grow. What is the probability that it eventually dies out?
Let \(q\) be the extinction probability. The outbreak dies out if the first case infects nobody, or infects two people whose separate outbreaks both die out:
$$q = \tfrac14 + \tfrac34\,q^2 \;\Longleftrightarrow\; 3q^2 - 4q + 1 = 0 \;\Longleftrightarrow\; (3q-1)(q-1) = 0.$$Branching-process theory says the extinction probability is the smallest root in \([0,1]\), so \(q = \tfrac13\). Even with \(R_0 > 1\), one outbreak in three disappears by chance. Try it in the outbreak game.
Counting the pieces of a virus
In Caspar–Klug theory, an icosahedral capsid with triangulation number \(T\) is built from \(60T\) identical proteins, grouped into 12 pentamers (5 proteins each) and some number of hexamers (6 proteins each). How many hexamers are there? How many for \(T = 3\)?
Count proteins two ways. If there are \(H\) hexamers,
$$60T = 5\cdot 12 + 6H \;\Longrightarrow\; H = 10(T-1).$$For \(T = 3\): \(180\) proteins, \(12\) pentamers and \(20\) hexamers. Allowed values are \(T = h^2 + hk + k^2\) for whole numbers \(h, k\), so \(T = 1, 3, 4, 7, 9, 12, 13, \dots\) See them in 3D in the virus shells game.
Outbreak simulator
Virus shells in 3D
Guess the \(R_0\)
Darts for \(\pi\)
Grow the hat tiling
Beat the forecaster
Drag in the shaded area to draw your forecast.
How far back should the model look?
Watch a network learn