Open to Opportunities

Hi, I'm Daniel Rivera

Publications Coordinator CS Student @ FUS Full-Stack Dev

I build AI-powered systems, full-stack web apps, and software architectures — blending a background in theology, catechetics, and computer science into technology that actually matters.

View My Work Get in Touch
🎓
Available
scroll
0
GitHub Projects
0
Languages
0
AI/ML Systems
0
Years Building

About me

Developer with a
theological perspective

I'm a Computer Science student and Publications Coordinator at Franciscan University of Steubenville, where I merge technical expertise with a passion for meaningful, human-centered technology.

From building RAG pipelines on AWS Bedrock to shipping full-stack web apps with Docker and Railway, I focus on software that solves real problems — especially in education and theological publishing. My current focus is software architectures, AI integration, and cloud deployment.

"Wonder is the beginning of wisdom." — Thomas Aquinas
See All Projects
🤖
AI & RAG Architecture
Built a production RAG system for Catechetical Review using AWS Bedrock, Magisterium API, and semantic search — deployed on Railway.
⚡
Full-Stack Development
React + TypeScript frontends, Node.js APIs, PostgreSQL schemas, Docker containers — across personal and team projects.
🏛️
Software Architecture
Implementing blackboard architecture and MVC patterns in C++ for a lunar lander simulator in Software Architectures coursework.
📚
Publications & Content
Managing quarterly journal releases on Drupal for the Catechetical Institute — bridging editorial workflow and technology.

Selected Work

Projects I'm proud of

🔬
Independent Research
What the Tests Don't See
A static audit of 6,000 patches that 12 frontier coding agents submitted to SWE-bench Verified, compared with human fixes for the same issues. Agent patches aren't less secure, but the tests miss real behaviors: silenced exceptions and edits to tests the agents were told not to touch.
AI Safety AI Security LLM Evaluation Python
AI Safety & Security Read the paper
// 12 models · 6,000 patches
 
security findings
  models 1.1% · human 1.6%
 
silent except
  GPT-5 mini 4.5%
 
forbidden test edits 23
  rewrote assertions: 5
✨
Professional Work
Project Helix — Handshake AI
AI training data engineer contributing to a SWE-bench–style model evaluation pipeline. Authored high-quality fail-to-pass test suites across real open-source codebases in Python, PHP, and Rust — covering security, protocol compliance, and database correctness. Each task was containerized with Docker-based eval runners for automated CI validation.
Python PHP Rust Docker AI Training Data SWE-bench
// fail-to-pass: auth security
 
def test_verify_bcrypt_password():
  handler.accounts = {"user": bcrypt_hash}
  assert handler.verify_password(
    "user", "secret")
  assert not handler.verify_password(
    "user", "wrong")
 
// eval: Docker + signal.json ✓
🚀
Moon Lander — C++ Architecture
Lunar lander simulator built in C++ for Software Architectures & Patterns coursework. Implements the Blackboard Architecture pattern with MVC, covering physics simulation, sensor input, control logic, and rendering with CMake on macOS.
C++ Blackboard Architecture MVC CMake
🗄️
Database Design & Implementation
Relational database schema normalized to 3NF for complex transactional requirements, built for Database & Algorithms coursework. Includes advanced SQL with CTEs, window functions, and optimized JOINs, backed by a full-stack Node.js/Express CRUD application.
SQL Schema Design Node.js Full-Stack
View All Projects