Security Administrator · Rowan University · New Jersey

I build the systems security teams actually use, and study the ones they can’t yet trust.

A decade across Sierra Leone, China and the U.S. in network engineering, Python development and security operations. Today I run vulnerability management and detection work at Rowan University, build the tools and apps I wish existed, and research indoor localization and AI risk.

FocusSecOps · Vuln. management
ResearchWi-Fi localization · AI risk
EducationMS Cybersecurity, Rowan
Portrait of Joseph Sahr Sellu
NowShipping Lumen Notes updates, building Aegiora, and finishing the Accord notes app. Indoor localization manuscript under review.
01 — Apps

Products I build and ship

Apps and platforms I design, build and run, from first sketch to store listing.

Lumen Notes screenshot 1Lumen Notes screenshot 2

Lumen Notes

Live on the App Store

A note app for Christians: the KJV Bible offline, sermon notes, a prayer journal with answered prayers, and focused study time that blocks distracting apps until you write what you learned. Free, no ads, and nothing leaves the phone for a cloud AI.

Platforms
iPhone · iPad (Android coming)
Built with
Expo, React Native, TypeScript, Supabase
Pricing
Free, optional tip jar
02 — Client & internal work

Tools built for real teams

Most of this runs inside organizations, so the public versions are sanitized case studies.

Featured case study

2024 – present
VulnTrack dashboard

VulnTrack: replacing the vulnerability spreadsheet

A tracking platform that imports Tenable findings, detects duplicates and resurfaced issues, and holds owners to remediation SLAs. Management gets a printable executive view instead of a pivot table.

Role
System designer and developer
Stack
Django, Python, SQLite, Tenable
Status
In use · sanitized case study
03 — Research

Research & publications

CG-AGC-WKNN
Manuscript under review

Confidence-gated adaptive geospatial-cell candidate generation for Wi-Fi fingerprint indoor localization

Weighted KNN over a full radio map is slow and brittle when signal strength drifts. CG-AGC-WKNN narrows the search to adaptive spatial cells and falls back to global WKNN only when the cell evidence is weak.

8.63%lower mean error than Global WKNN (9.181 m vs 10.048 m)
12.33%lower median error
77.42%smaller online candidate search space
Dataset
UJIIndoorLoc (multi-building, multi-floor)
Method
Adaptive cells + local WKNN + confidence gate
Metrics
Mean, median and P90 error; candidate reduction
Next
Adaptive cell expansion, live indoor testbed
Policy
IoT cyber law and security governanceManufacturer liability, privacy, consent and international coordination
Working paper
Applied
Modernizing vulnerability managementSLA monitoring, ownership and remediation accountability
Practice
Applied
Automation in security operationsKQL, PowerShell and Python to cut manual analyst work
Practice
Computer vision
Drone imagery for beach trash detectionDetecting visible trash from aerial images for environmental monitoring
Concept
05 — Experience

Roles

Security AdministratorRowan University · 2024 – present
Security Analyst InternRowan University · 2022 – 2024
Python ProgrammerOpenPath LLC · 2018 – 2022
IT Systems EngineerInsight Technologies (SL) Ltd · 2014 – 2020

Education & credentials

MS, CybersecurityRowan University · 2022 – 2024
Graduate studies, Information & Communication EngineeringHuazhong University of Science & Technology · 2018 – 2020
BSc (Hons), Computer ScienceNjala University · 2010 – 2015
SSCP · ISC2 CC · CCNA · Microsoft SC-200 (training) · Microsoft Sentinel Ninja · Google Python Automation · AWS Cloud Practitioner Essentials