Computer Science and Mathematics at Southern Miss

SuramyaAngdembay

I'm an undergraduate researcher at The University of Southern Mississippi, where I study Computer Science and Mathematics. I work as a Research Assistant at the Institute for Advanced Analytics and Society, contributing to AQUAVIEW through ocean data workflows, MCP connector work, and ML-corrected mixed layer depth and hurricane heat potential prototypes.

I was awarded the $2,000 Wright W. & Annie Rea Cross Mathematics Research Scholarship in 2025, and I am part of the USM Honors College through the Honors Keystone Scholar Award. In 2026, I was also selected for the Break Through Tech AI Fellowship by Cornell Tech.

My work is project-first: applied ML, scientific computing, HPC workflows, ocean and weather data systems, and research prototypes that turn messy real-world data into something measurable, explainable, and useful.

Resume GitHub LinkedIn

17 public GitHub repos

23 eBay ML public rank

2 conference talks

Applied ML NER, anomaly detection, forecasting

Scientific Systems HPC, Slurm, ocean data, APIs

Data Products Dashboards, mobile apps, geospatial workflows

Featured mention

COMAP highlighted our first-time MCM team story.

After our 2025 Mathematical Contest in Modeling project on Olympic medal prediction, COMAP shared the story of our team and how we approached the contest as first-time participants.

Read the COMAP feature

Selected work

Projects

Research 2024 - present

Insider Threat Detection

Unsupervised anomaly detection pipeline for CMU CERT insider behavior data using Featuretools DFS, PCA dimensionality reduction, and Gaussian Mixture Models.

Built feature synthesis over email, HTTP, device, file, logon, and psychometric logs.
Used PCA to retain dominant behavioral variance before density-based anomaly scoring.
Ran large processing workflows on USM's Magnolia HPC cluster with Slurm.

Python Featuretools PCA GMM HPC

View repository

Challenge 2025

eBay University ML Challenge

End-to-end German eBay Motors NER pipeline that extracts structured product attributes from short listing titles.

Domain-adaptive MLM pretraining on 2M unlabeled listing titles.
Fine-tuned XLM-RoBERTa with BIO token labels for aspect extraction.
Generated EvalAI-ready submissions and reached Rank 23 on the public leaderboard.

PyTorch Transformers NLP XLM-RoBERTa

View repository

AQUAVIEW 2026

MLD ML Recomputation

Historical replay prototype for machine-learning-corrected mixed layer depth in the Gulf of Mexico, combining RTOFS model fields with nearby in-situ observations and provenance-aware confidence output.

Built a replay sandbox for querying corrected MLD at a clicked map location and date.
Integrated model output, nearby observations, correction fields, confidence, and provenance.

Python FastAPI React RTOFS Argo/WOD ML correction

View repository

AQUAVIEW 2026

HHP ML Recomputation

ML-corrected Hurricane Heat Potential replay sandbox for Gulf hurricane cases, applying TEOS-10 thermodynamics to Argo and RTOFS profiles before correcting scalar TCHP estimates.

Computed TCHP/HHP with TEOS-10-backed profile processing for storm replay cases.
Benchmarked tabular ML corrections against raw RTOFS using Argo-derived targets.

Python TEOS-10 GSW FastAPI RTOFS Random Forest

View repository

Quantum 2025

Quantum Galton Board Simulator

Qiskit simulator for quantum walks that maps a Galton board to reusable n-layer circuits and compares quantum, classical, and biased distributions.

Qiskit Python Simulation Metrics

View repository

API 2025

AireChemistry Weather API

FastAPI-backed South Africa weather and air-quality service with gridded fields, point visualizations, GeoJSON industrial layers, and NumPy data pipelines.

FastAPI NumPy GeoJSON Weather data

View repository

Mobile ML 2024 - 2025

Wildfire Spread Prediction App

CNN-based wildfire spread predictor paired with a React Native map interface, Firebase notifications, and FastAPI model serving on AWS EC2.

TensorFlow React Native FastAPI Firebase

View repository

Modeling 2025

Olympic Medal Prediction

First-time COMAP MCM project for 2028 Olympic medal prediction. The team tested linear regression, ARIMAX, Bayesian and grey models, then selected Random Forest regression for the final pipeline.

Python Regression Random Forest ARIMAX COMAP MCM

View repository Read COMAP feature

Conferences and talks

Presentations

Mar 2025

AMS Southeastern Sectional Meeting

Presented "Insider threat detection using Principal Component Analysis" in the Undergraduate Mathematics and Statistics Research III special session at Clemson University.

AMS talk listing

Nov 2025

SC25, The International Conference for High Performance Computing

Selected for the ACM SIGHPC Immersion Program, completed an OLCF HPC/AI crash course, and gave a Student HQ lightning talk on the SIGHPCI experience.

Jul 2025

CompuCell3D Summer School and Hackathon

Accepted into Indiana University's virtual summer school and hackathon on multicell virtual tissue modeling.

Jun - Jul 2025

WISER and Womanium Quantum Summer Program

Participated in the quantum computing summer program and built the Quantum Galton Board project with a final presentation deck.

Research and applied work

Experience

Undergraduate Research Assistant

University of Southern Mississippi | Aug 2024 - present

Researching PCA-based dimensionality reduction and unsupervised anomaly detection on insider threat behavior data, with parallel workflows on Magnolia HPC.

Research Assistant

Institute for Advanced Analytics and Society, USM | Jan 2026 - present

Contributing to AQUAVIEW under Dr. Gregg Jacobs, including MCP connector work, mixed layer depth recomputation, and ML-corrected hurricane heat potential workflows.

AireChemistry Weather API Developer

AireChemistry | Jul 2025 - Dec 2025

Designed FastAPI endpoints, data arrays, map textures, and point-based weather and air-quality visualizations for South Africa.

Mathematical Modeling and Optimization Intern

Optimal Answers LLC | Jan 2026 - present

Building structured project data and optimization rules for Gulf Coast Restoration Fund allocation models.