St. Louis, Missouri · 38.6°N 90.2°W GPA 4.00 / 4.00 — 2026

Addison
Thurston.

Computer science student at Saint Louis University and software engineer specializing in security for agentic and machine-learning systems. Thorough, curious, and committed to tools that genuinely help people.

StatusB.S. Computer Science · May 2027
FocusAgentic Security · ML Integrity · HPC
Emailaddithurston@gmail.com
GitHub/AddiThurston
LinkedIn/addison-thurston
ResumeDownload .docx
01 / About
— a short portrait
Hello Addison Thurston
Addison Thurston St. Louis · 2026

I'm a computer science student and software engineer focused on security for agentic and machine-learning systems. I build identity verification, token enforcement, and demonstration tooling at an early-stage agentic security startup, and I lead an open-source project on model integrity.

Lately that means hardening Auth51 at Unforge, shipping an MCP server for backdoor-detection defenses with Mithridatium, and teaching data structures in C and C++. I like tools that are thoughtful under the hood and honest at the surface.

Outside of code: School of Science and Engineering ambassador, Girls Who Code executive board, SLU Pep Band, and the occasional competitive programming round.

4.00/4.00
Cumulative GPA
1st
HackSLU Education Track
4×
Dean's List
02 / Skills
— the working toolkit

A toolkit built on fundamentals, sharpened on real projects.

01
Languages
C · C++ · C# · Java · Python
02
Core CS
Data Structures · Algorithms · OOP
03
Security
OAuth 2.0 · JWT · DPoP · API Security
04
Agentic AI
LangChain · LangGraph · MCP
05
Machine Learning
PyTorch · Model Integrity · Computer Security
06
Web
React · Vite · FastAPI
07
Engineering
Git · CI/CD · Docker
08
HPC
PaRSEC · Python/Julia Bindings
03 / Experience
— research, teaching & industry

Where the work happens.

May 2026 → Now

Agentic System Security Engineer

Unforge Inc. · St. Louis, MO
  • Ported and hardened Auth51, a Python and FastAPI library that validates JWT signatures, issuer, audience, scopes, and DPoP bindings before an AI agent's API call can run.
  • Authored a 40-case pytest suite and CI pipeline (coverage, Black, mypy, Bandit) covering token replay, scope mismatch, and audience confusion.
  • Built live demonstrations of agent identity and intent-binding enforcement, plus a reusable scenario template for new demos.
  • Presented MCP internally, surfaced a gap in tool-call coverage, and co-designed the proxy architecture adopted to close it.
  • Prototyped a LangGraph agent instrumented with the Auth51 SDK to validate agent fingerprinting and intent-token propagation.
  • Authored public product documentation and rebuilt the company site in React and Vite, including sitemap and robots.txt for search indexing.
FastAPIJWTDPoPLangGraphMCPReact
Aug 2026 → Now

Tech Lead, Mithridatium

Open Source with SLU · St. Louis, MO
  • Lead a team of four on Mithridatium, an open-source Python framework that detects backdoors and data poisoning in pretrained image classifiers using four published defenses (MMBD, FreeEagle, STRIP, and AEVA).
  • Own the roadmap and team process — semester targets, working agreement, and weekly faculty-client updates.
  • Built the project's first GitHub Actions CI pipeline, gating every pull request on 67 tests across 17 files.
  • Review contributor PRs and triage the backlog, turning research goals into scoped tasks with acceptance criteria.
PythonPyTorchCI/CDLeadershipOpen Source
Jun 2026 → Now

Undergraduate Researcher

T-Lab · Saint Louis University
  • Build an agentic system that lets an LLM select, run, and interpret backdoor-detection defenses against untrusted pretrained models.
  • Developed a Python MCP server with FastMCP exposing four ML defenses as agent-callable tools, returning schema-validated JSON for PyTorch and Hugging Face classifiers.
  • Designed a LangGraph supervisor with a capability registry that routes a target model to the right defense based on weight access, dataset availability, and architecture.
  • Develop a LangChain intrusion-detection agent for AI pipelines on SLU's Libra HPC cluster, integrating the toolkit as a versioned Git submodule.
MCPLangGraphLangChainPyTorchHPC
Aug 2026 → Now

SSE Ambassador

Saint Louis University · School of Science and Engineering
  • Represent the school to prospective students and families — campus and laboratory tours, plus student panels at open houses.
  • Create social and marketing content that highlights student research and academic programs.
OutreachCommunications
Sep 2025 → Now

Research Assistant

Saint Louis University · HPC Lab
  • Collect and analyze data for the Parallel Runtime Scheduling and Execution Controller (PaRSEC) library.
  • Expose PaRSEC's C APIs to Python and Julia, extending HPC reach beyond the native C ecosystem.
  • Co-author a paper with graduate researchers on the merits and performance of PaRSEC for Python.
CPythonJuliaHPCResearch
Aug 2025 → Now

Teaching Assistant

Saint Louis University · CS Dept.
  • Instruct and mentor 30–40 students in data structures and computer systems (C/C++).
  • Monitor student progress and relay timely feedback to professors.
CC++TeachingMentorship
04 / Selected Projects
— three things I've built

Tools that are thorough, useful, and honest.

Hackathon · First Place, Education Track 2026 · HackSLU

CanvasCurator

A Python web app that turns uploaded images or text prompts into fully-structured Canvas assignments and quizzes. Under the hood: a three-stage pipeline — Google Cloud Vision extracts text via OCR, Gemini 2.5 Pro structures it into coursework, and the Canvas API publishes it directly to a live course. Cuts manual formatting time dramatically for educators.

PythonGoogle Cloud VisionGemini 2.5 ProCanvas APIFlask
AI Web Application2026

InSiteful Mind

An AI-driven web app that analyzes user-provided data to surface meaningful insights and support better decisions. Designed around clean software-engineering principles — modular data processing, intelligent analysis, and a presentation layer that keeps results readable and actionable.

PythonWeb AppData AnalysisAI/ML
Team Project · Secure Systems2025

Secure HealthComm

A secure messaging platform for patients and healthcare providers, built in Java with Swing. Implemented end-to-end encryption to guarantee confidentiality and integrity of sensitive health data. Collaborated with Git — branches, merges, and real conflict resolution under tight deadlines.

JavaSwingE2E EncryptionGit
05 / Awards & Activities
— on the side & on the record

◆ Honors & Awards

  • 1st Place · HackSLU Education Track2026
  • Boeing BOLD Scholarship2024–
  • Dean's ListF24 · S25 · F25 · S26
  • SLU Honors Program2023–

◆ Activities

  • Girls Who Code · Executive BoardOngoing
  • HackSLU Hackathons2025 · 2026
  • ICPC · Programming CompetitionOngoing
  • SLU Pep BandOngoing
06 / Contact
— let's talk

Let's build
something useful.

addithurston@gmail.com ↗
Phone
314 · 327 · 1837
Located
St. Louis, MO
GitHub
AddiThurston
LinkedIn
addison-thurston
Resume
Download .docx