Skip to content

10 projects

Work that shipped!

In progress

02
01AI / ML2026In progress

Adaptive AI Educator

Building the podcast format for a multi-format learning platform

A learning platform that teaches aquatic invasive species through several formats at once, published at HCII 2026. I am building the podcast side of the implementation: scripted episodes voiced with ElevenLabs, presenters modeled in Blender, and the whole thing rendered in the browser through Three.js. The research is peer-reviewed and the build is underway.

ElevenLabsBlenderThree.jsTypeScriptWebGL
HCII
Published 2026
02XR2026In progress

Multi-User VR Classroom

A shared classroom for Meta Quest, built around lecture, challenge, and debrief

The VR learning platform I build as a research assistant on the NSF project. Students join a shared classroom from a Meta Quest headset. A 50-minute class runs as three or four cycles of a lecture video, a challenge on that topic, and a debrief for questions. Row-level security keeps the answer key in the database and off the client. It ran in a two-week class playtest with six students in November 2026.

TypeScriptThree.jsIWSDKWebXRPostgreSQL+4
11
RPC functions

Shipped

01
02Web2026Live

Temple Display Board

A live event board and daily notification job for a community temple

A fullscreen display for the Hindu Temple of Eastern Shore in Salisbury, running unattended on a TV in the building. It shows the current month's events and the people sponsoring them, and it updates the moment the data changes. Behind it, a scheduled job checks each morning for anything happening that day and reports what it did.

Next.js 16TypeScriptReact 19Tailwind CSSshadcn/ui+6
Live
Realtime updates

Everything else

07
04Security2026

LLM Agent for Threat Hunting

A local language model that triages a live security alert stream

A threat-hunting agent that builds a running profile of each user from a live Wazuh alert stream, then prompts a locally hosted Qwen2.5-7B with that frequency and business-hours context. The reply is constrained to JSON carrying severity, confidence, and a recommended response action, so a SOC pipeline can route it without a human reading prose first.

PythonOllamaQwen2.5-7BWazuhJSON Schema
100%
Local inference
05AI / MLSpring 2026

Insider Threat Detection from Windows File-Share Events

Unsupervised ML on 4656/5145 telemetry in a live Zero Trust lab

Windows domains emit file-access telemetry constantly. Events 4656 and 5145 record who touched what, from where, and whether access was granted, and almost nobody looks at it. This work builds an unsupervised detection pipeline on that stream inside a controlled Active Directory lab, folding raw events into behavioral sessions and scoring them with Isolation Forest, LOF, and One-Class SVM.

Pythonscikit-learnSHAPWindows Server 2022Active Directory+1
18,572
Events collected
06SystemsMar 2025

GPU Acceleration for Cryptanalysis

AES-128 key search, CUDA against OpenMP, measured

A direct measurement of what wide parallelism does to brute-force cryptanalysis. The same AES-128 key search was written twice, once with OpenMP across 8 CPU threads and once in CUDA across 40,960 threads, then run over a 4.29 billion key space. CUDA finished in 123.5 seconds against OpenMP's 3,928, a 32x gap on a laptop GTX 1050.

CUDAOpenMPC / C++AES-128
32x
Speedup
07SecurityMar 2026

Campus Network Security Design

Segmentation, access control, and detection for a three-floor academic building

A full network security design for an academic building at UMES: offices, conference rooms, classrooms, teaching labs, and a server room across three floors. Six VLANs partition the network by function, an ACL matrix governs what each one may reach, and a SIEM plan turns the resulting traffic into alerts mapped to MITRE ATT&CK techniques. This is a design study rather than a deployment. What follows are the decisions and the reasoning behind them.

Cisco IOS802.1Q VLANs802.1X / RADIUSExtended ACLsWPA3-Enterprise+5
6
VLANs
08SecurityDec 2025

Zero Trust Architecture over SDN

Software-defined enforcement with live behavior profiling

A working Zero Trust implementation built on software-defined networking with Mininet and a Ryu controller. Every packet passes three enforcement stages: ICMP rate filtering, payload content inspection, and protocol allow-listing. A profiling layer tracks per-host behavior over time. Tested against staged flood attacks, the system isolated the offending host as its share of traffic climbed past 56%.

PythonMininetRyu controllerOpenFlowMatplotlib
3
Enforcement stages
09SystemsMay 2025

Where GPU Offload Stops Paying Off

A 90-configuration sorting benchmark across CPU, OpenMP, and CUDA

Merge sort and quick sort measured across sequential CPU, OpenMP, and CUDA, over sorted, random, and reverse-sorted inputs at 2, 4, and 8 million elements, on both Windows and macOS. The GPU lost every single run. That negative result is the useful part, because it marks the boundary where transfer cost and irregular memory access beat raw thread count.

C / C++OpenMPCUDAWindows 11macOS
90
Configurations
10SystemsNov 2025

Agile vs Waterfall, Measured on Real Builds

Two network simulators, two methodologies, one comparison

Instead of arguing methodology in the abstract, this study built two working Mininet-WiFi simulators under different process models. A satellite orbit simulator was developed Agile across four sub-goals. A concurrent network performance tester was developed Waterfall through full SRS and design documents. The resulting code structures diverged in ways the process literature predicts but rarely demonstrates.

PythonMininet-WiFiiperfMatplotlib
2
Full simulators