v₀ · Chattogram, Bangladesh · 22.35°N, 91.78°E

Ifteha finds patterns in the noise.

Computer Science & Engineering graduate of IIUC and a machine-learning researcher in the making. Constellations, after all, are just graphs — and she's learning to read them.

Read the research ↓ GitHub ↗

v₁ · origin node

Every graph starts with a first vertex.

It started small: one line of code that actually worked, and the little thrill of making a computer do something. That grew into a B.Sc. in Computer Science & Engineering at the International Islamic University Chittagong — four years of late nights, stubborn bugs, and slowly falling for the quieter corners of CS: data, theory, and how things fit together. She's still learning. Honestly, that's her favorite part.

degree
B.Sc. in CSEIIUC, Chattogram
status
Graduated · 2026
focus
ML · Graphs · Databases
thesis
Spammer Group IdentificationHeterogeneous VGAE

v₂ · the toolkit

Stars she navigates by.

A researcher's stack, built patiently: languages for thinking precisely, databases for keeping the world in order, and machine learning for finding what hides between the rows.

✶ LANGUAGES

Thinking in code

From pointer arithmetic to list comprehensions — comfortable across paradigms, from bare-metal to high-level.

CC#Python
✶ DATABASES

Order from chaos

Database management systems — schema design, normalization, and queries that respect both the data and whoever reads it next.

DBMSSQL
✶ MACHINE LEARNING

Learning from data

Her home ground: representation learning, graph neural networks, and models that spot what humans miss.

MLGraph ModelsVGAE

v₃ · the anomaly

Some nodes don't belong to the constellation.

Online platforms are graphs of people. Most nodes behave like stars — steady, predictable. But spammers move in coordinated packs, at suspicious speeds. Her undergraduate dissertation taught a machine to see them.

“Spammer Group Identification Using Heterogeneous Variational Graph Autoencoder with Velocity-Adaptive Temporal Modeling”

HETEROGENEOUS

Real networks aren't one kind of thing — users, posts, reviews, timestamps — all different node types, all in one graph.

VGAE

A variational graph autoencoder compresses the network into its essence, then tries to rebuild it. What can't be rebuilt cleanly is suspicious.

VELOCITY-ADAPTIVE

Spam has a tempo. The model adapts to how fast behavior changes over time, catching bursts of coordinated activity.

RESULT

Not just single spammers — whole spammer groups, identified by how they cluster, move, and act together.

FLAGGED: GROUP ANOMALY
fig. 01 — the network as the model sees it legitimate users coordinated spammer group

v₄ · the build log

Written in commits.

Coursework, experiments, and things built at 2 a.m. because the idea wouldn't wait — all of it lives on GitHub, in the open.

RESEARCH CODE

Dissertation experiments

Graph autoencoder pipelines, temporal modeling, and spammer-group detection benchmarks.

view on GitHub →
ML NOTEBOOKS

Machine learning work

Models, training runs, and the notebooks where ideas get tested against real data.

view on GitHub →
COURSEWORK & MORE

Everything else

C, C#, and Python projects from four years of CSE — database systems included.

browse all repos →

v₅ · off the clock

Even models need rest states.

A well-trained researcher knows when to stop training. Off the clock, she's fiercely committed to two disciplines of her own.

Professional-grade sleep

Some people optimize hyperparameters. She also optimizes REM cycles. The best debugging happens after eight hours — this is science, allegedly.

Collecting coordinates

Travel is graph traversal in the physical world: new nodes, new edges, new data. The itinerary list is long, and the visited-set keeps growing.

v₆ · the next edge

The graph isn't finished. There's more to learn.

Degree in hand, and hoping to keep growing as a researcher — good problems, good mentors, and work worth losing a little sleep over (only a little — she guards her sleep). Always happy to talk about research, ideas, or anything graph-shaped.

eftehabinte380@gmail.com
GitHub ↗ Say hello ✉