Projects
Research & hands-on work
Selected work from my Master’s in Cybersecurity at the University of Central Florida and earlier hands-on infrastructure labs.
Comparing LLM-generated and human-identified threats in software threat modeling
Can large language models help security teams threat-model software? Our group runs locally hosted models against the same systems human analysts reviewed and compares coverage, accuracy and false positives. I lead the evaluation design and analysis.
VPN vs. SSH tunneling: performance and security benchmark
A reproducible benchmark comparing WireGuard, OpenVPN and SSH tunnels on throughput, latency and security properties, to help decide which tool fits which remote-access scenario.
Can a language model plan, or is it copying plans it has seen?
A critical review of evidence on LLM planning, including PlanBench and obfuscated “Mystery Blocksworld” tasks, and what it means for trusting AI agents with multi-step decisions.
Reinforcement learning in MiniGrid DoorKey
Investigates why a PPO baseline solves the DoorKey 5×5 and 6×6 environments but fails at 8×8, using a three-group design: control, milestone reward shaping and full observability.
Infrastructure & security labs
Building a secure intranet
Video walkthrough of a proof-of-concept intranet published through a reverse proxy.
Designing a LAN
A complete LAN design project, from requirements to topology.
Hybrid-cloud user management
Identity and access management across on-premises and cloud environments.
Securing shared folders on Windows
Basic LAN design and access control for Windows shared folders.