Web · AI · 2026
Trefelle
A practice space for engineering judgment that runs AI and real code entirely in your browser. Your keys never leave your device.
Software engineering student · Temple University
Netcode, developer tools, and AI evaluation.
I build multiplayer netcode, developer tools, and evaluation pipelines for AI systems. Every project below has a demo you can use.
i
Netcode that holds up under lag
Combat validated against tick-indexed server snapshots, with rate-limiting and lag compensation benchmarked under simulated ping.
ii
Evaluations that catch real failures
Docker-run model evaluations on real repositories, structured failure-mode analysis, and datasets with planted traps.
iii
Tools for people who build
Arboryn merges Roblox object trees, and Trefelle gives engineers a practice space with real code execution and bring-your-own AI.
01Selected work
Projects
Roblox · Netcode · 2024 - Present
Aspirer's Workshop
A Roblox game with 3,000+ visits. I built the combat systems and the client-server networking behind them, including parry fairness across 0 to 400ms of ping.
Developer tools · In progress
Arboryn
A local-first three-way merge engine for Roblox object trees. Currently in Phase 0, a Studio capture harness.
AI evaluation
VLM Image-Pair Judge
A Python tool that uses vision-language models to judge pairs of AI-generated images against a written rubric.
02Where I've worked
Experience
Data Annotation Tech
AI Model Evaluator, Software Engineering · Jul 2026 - Present · Remote
- Configured Docker environments to run AI models against open-source repositories, enabling reproducible evaluation across codebases.
- Ran model inference and captured execution traces to compare outputs against expected behavior.
- Documented model errors with structured failure-mode analysis and recommended corrections.
- Built evaluation datasets and grading rubrics for AI agents, including multi-file synthetic corpora with planted traps across finance, legal, and engineering domains.
Handshake AI
Data Annotator, contract · Nov 2025 - Present · Remote
- Labeled and rated AI model outputs, including images, against task guidelines.
- Reviewed AI-generated code and documented issues to support model improvement.
Concilio
Digital Media Specialist · Jun - Sep 2023 · Philadelphia, PA
- Mentored digital media cohorts across 2 summers and trained users on operating systems and networks.
- Resolved escalated technical issues through proactive troubleshooting to minimize downtime.
Johnson Real Estate
Apprentice · Jun 2022 - Dec 2024
- Managed multiple property portfolios and client communication.
03Background
About
I'm a computer science student at Temple University in Philadelphia. At work, I test how AI models fail on real code and turn those failures into datasets and rubrics.
Temple University · B.S. Computer Science, 78 credits completed, 14 in progressCommunity College of Philadelphia · A.S. Computer Science
Data Structures and AlgorithmsSystems Programming (C)Operating SystemsProbability and Statistics
Languages
PythonSQLTypeScriptJavaScriptHTML/CSSCC++C#Lua/Luau
AI and evaluation
LLM and agent evaluationRubric designAdversarial datasetsFailure-mode analysis
Engineering
Git/GitHubDockerLinuxNode.jsSupabaseVercelREST APIsClient-server architectureDebugging