Gunnar Von Bergen

Software Engineer & AI/ML Researcher

GitHub LinkedIn Contact

Snapshot

  • B.S. Computer Science (Honors) — Magna Cum Laude
  • Miami University • GPA 3.88
  • Minor in Japanese
  • Global Readiness Certificate
  • AI Research + Software Engineering

Research

Research papers and ongoing projects.

Med-CoT

Paper

Accepted — IEEE Big Data 2025 (Macau)

Proactive clinical support agent that monitors doctor/patient dialogues and intervenes on confusion, safety risk, or missing information; led the chain-of-thought structure design with SOAP-note formatting and a decision space policy.

Clinical AILLM ReasoningEvaluation

Historia Augusta Authorship

In progress

Exploratory research

Early-stage study using ML and GenAI to investigate the true author(s) of Historia Augusta, an unresolved attribution problem with limited ground truth.

Authorship AttributionML/NLPExploratory

Featured Projects

Software engineering projects across full-stack and product builds.

All repos ↗

Slope Search

live code

Full‑stack ski area search across 4,500+ areas with dynamic filters. React + Express + SQL.

ReactExpressMySQLDocker

Quick Quit (Chrome Extension)

code

Privacy tool that clears history and inserts decoys on partner sites via JS triggers.

ExtensionNodeAWS EC2

About

I’m a software engineer and AI/ML researcher who learns by building and breaking real systems. Most recently at Cleveland Clinic’s Discovery Accelerator, I explored multi-agent LLM systems for automating bioinformatics workflows, refactored their tooling through an MCP layer, and spent a lot of time understanding where these agents actually fail. I care about clear technical communication, picking up new topics quickly, and being straightforward about what a system has and hasn’t accomplished.

  • Java, Python, JS/TS, C++
  • React, Node/Express
  • Postgres/MySQL, SQL
  • Docker, GitHub Actions

Recent Experience

Research Intern — Cleveland Clinic

Cleveland, OH • May 2026 - Aug 2026

Selected for the Discovery Accelerator, Cleveland Clinic's biomedical research partnership with IBM.

  • Independently drove a research project on multi-agent LLM systems for automating bioinformatics workflows on the Galaxy platform, running teams of up to four local 70B models across four GPUs
  • Drove, tested, and validated a refactor routing the framework's built-in tools through an MCP layer, bridging asynchronous MCP operations into a synchronous system for more testable, controllable tool exposure
  • Enabled agent teams to source Zenodo data, navigate Galaxy histories, and explore and call from thousands of platform tools with mixed reliability
  • Identified and documented core failure modes including unverified claims between agents, premature task completion, supervisor agents overriding teammates, and context bloat from tool schemas, prioritizing honest evaluation over inflated end-to-end results
  • Researched and delivered a byte-pair encoding tokenization talk to program interns and mentors, hand-working examples and compiling glitch tokens