AI & Agent Projects
Production and hackathon AI systems — tool-calling LLM agents, custom MCP servers, and LLM evaluation frameworks.
Crypto Risk Analysis Agent
Production Agents Hackathon
A full-stack AI agent that evaluates token and wallet risk by combining blockchain data from Helius, Moralis, DexScreener, and TRM Labs via the Anthropic API.
- Structured prompts and tool-calling workflows convert noisy API inputs into grounded risk scores and interpretable recommendations.
- A custom MCP server exposes four blockchain-analysis tools over HTTP, SSE, and stdio transports.
- Integrated with a Node.js/Express backend and deployed with Docker, retries, rate-limit handling, and health checks.
Dynamic Benchmarking of LLMs
SFSU — Computational Linguistics
A comparative evaluation framework for static, dynamic, and metalinguistic LLM benchmarks, assessing whether benchmark scores reflect genuine language understanding.
- Focuses on benchmark contamination risk, reasoning quality, and benchmark reliability.
- Analyzes model performance and failure patterns across evaluation methods.
Additional Hackathon Awards
- REAL Agent Challenge — AGI Inc. × OpenAI × Lovable Hackathon 2nd Place
- Frontegg Sponsor Prize — Self-Evolving Agents Hackathon 1st Place
Software & Infrastructure
Production services, hosting infrastructure, and web platforms I build and operate.
QubeMC
A Minecraft Gaming and Hosting company I founded and operate. We offer various game modes and server hosting for both Bedrock and Java editions, with full-stack web apps, backend APIs, and automation behind the scenes.
View Project2B2TMCPE
A Minecraft Bedrock Edition anarchy server based on 2b2t.org. It's a chaotic environment where anything goes — backed by Linux infrastructure I administer.
View ProjectPersonal Website
This website is itself a project — a professional portfolio built with Bootstrap, and continuously refined as my work evolves.
View ProjectResearch
Linguistics and computational research presented at academic venues.
AI-Assisted Text Simplification
ECOLE Lab, San Francisco State University — Sep 2024 to Present
Investigating how large language models simplify complex texts and how AI-generated simplifications differ from human language processing in meaning, linguistic structure, and readability. Building Python-based NLP, evaluation, and data-analysis workflows.
Production of English /r/ by Prosodic Position
University of California, Santa Cruz — Sep 2022 to Jun 2023 · LURC 4, Jun 2023
Built Python workflows automating experimental trials and synchronizing multimodal speech data. Integrated DeepLabCut and Articulate Assistant Advanced for CNN-based tongue-contour tracking, with Whisper and LLM-assisted post-processing for transcription and labeling. Presented as a poster at UCSC's Linguistics Undergraduate Research Conference (LURC 4).
View Poster