---
title: "Work | Simon Rüba"
description: "Simon Rüba: LLM research and engineering, co-founder of enrik.ai. Kaggle AI Agent Security gold medal, 9th place, reasoning experiments and AI tools."
canonical_url: "https://simonrueba.com/"
md_url: "https://simonrueba.com/work.md"
last_updated: "2026-09-09"
language: "en"
---

# Simon Rüba — Work

LLM research & engineering · Innsbruck, Austria

Co-founder of [enrik.ai](https://enrik.ai/), an AI-native go-to-market (GTM) platform.

My research interests are LLM training, evaluation, and reliable reasoning. My public work includes language-model experiments, AI agent security, and tools for coding agents.

## Selected research

### [Kaggle AI Agent Security](https://github.com/simonrueba/ai-agent-security-9th-place)

I won a gold medal and placed 9th out of 4,211 in the Multi-Step Tool Attacks competition. I collaborated with Refractal AI, a London-based AI security startup, on the competition.

The public solution includes a technical working note and evaluation tooling for the submitted attack, reference defenses, and candidate search. The write-up documents the agent-assisted workflow, attribution, and limits of local replay.

- [Competition](https://www.kaggle.com/competitions/ai-agent-security-multi-step-tool-attacks/overview)
- [Technical write-up](https://github.com/simonrueba/ai-agent-security-9th-place/blob/main/WORKING_NOTE.md)
- [Refractal AI](https://refractal-ai.com/)

### [Ledger-of-Thought](https://github.com/simonrueba/cot)

Independent experiment · July 2026

I evaluated Qwen2.5-7B-Instruct on a bookkeeping task with reasoning depths of 4, 8, 16, and 32, using 32 tasks per condition and depth. The experiment compares chain-of-thought prompting with an external state ledger and ablations of its checking mechanism.

At depth 32, the ledger condition reached 50.0% accuracy versus 28.1% for the chain-of-thought baseline. Removing the checking mechanism produced the same accuracy at every tested depth: the checks added no measured accuracy in Stage A.

These preliminary results cover a single model and task family. This is an inference-time evaluation using a fixed pretrained model. The report includes the experimental setup, confidence intervals, ablations, and limitations; code and recorded results are public.

- [Stage A report](https://github.com/simonrueba/cot/blob/main/report_stageA.md)
- [Experiment code](https://github.com/simonrueba/cot/blob/main/cot.py)
- [Recorded results](https://github.com/simonrueba/cot/blob/main/lotA.json)

## Tools

- [Ariadne](https://github.com/simonrueba/ariadne): Code indexing & dependency analysis for agents
- [Tripwire](https://github.com/simonrueba/tripwire): Repository context for coding agents
- [BPE Visualizer](https://bpe-visualizer.com/): Interactive tokenization

- [BPE Visualizer source](https://github.com/simonrueba/bpe-visualization)
- [More on GitHub](https://github.com/simonrueba)

## Sitemap

- [Work](https://simonrueba.com/work.md): Simon Rüba: LLM research and engineering, co-founder of enrik.ai. Kaggle AI Agent Security gold medal, 9th place, reasoning experiments and AI tools.
- [Background](https://simonrueba.com/about.md): About Simon Rüba: co-founder of enrik.ai, previous engineering experience, education and skills. MSc Computer Science in progress. Innsbruck, Austria.
- [Contact](https://simonrueba.com/contact.md): Contact Simon Rüba in Innsbruck, Austria, about research engineering, LLM evaluation, enrik.ai, or his public projects.
- [Privacy](https://simonrueba.com/privacy.md): How Simon Rüba's portfolio handles website requests, contact messages, spam protection, retention, and privacy enquiries.

[Markdown sitemap](https://simonrueba.com/sitemap.md) · [Full portfolio](https://simonrueba.com/llms-full.txt) · [Agent guide](https://simonrueba.com/AGENTS.md) · [Terminology](https://simonrueba.com/glossary.md)