I build AI systems and I break things to make them safer.
AI engineer and security researcher based in Reykjavík, Iceland. Born and raised in Lagos, Nigeria, I moved to Iceland in 2023 and earned an MSc in Computer Science at the University of Iceland. Before that I spent five years working across AI and cybersecurity.
My work sits on both sides of the attack and defense line: I fine-tune and deploy language models, and I find and responsibly disclose vulnerabilities in real systems.
"He will cover you with his feather, and under his wings you will find refuge." 🎭
- 🍎 Apple Security acknowledgements on macOS 27 (App Intents and DataAccess) — support.apple.com
- 🏆 GKI 2026 university champion — Iceland's national AI championship, team CyboTRIX
- 🥈 Nordic AI Cup 2026 — Silver at the Nordic final in Reykjavík
- 📄 Published honeypot research — SBASH, FLLM 2025 (IEEE)
- 🤖 Icelandic LLM fine-tunes on Hugging Face, evaluated on the Miðeind Icelandic leaderboard
| Work | Description | Links |
|---|---|---|
| SBASH | A framework for designing and evaluating RAG vs. prompt-tuned LLM honeypots. Shell-based honeypots driven by small, locally run language models. | arXiv · DOI · IRIS |
| MSc Thesis | Computer Science, University of Iceland. | Skemman |
Co-authored with Helmut Neukirchen and Thomas Welsh, University of Iceland.
| Project | Description | Stack |
|---|---|---|
| VegaLingo | Icelandic language-learning app. Founder and developer. | React, Node.js, LLMs |
| Icelandic LLM fine-tunes | Gemma 3 and 4 models fine-tuned for Icelandic, evaluated on the Miðeind leaderboard. | PyTorch, Hugging Face, LoRA/SFT |
| GKI 2026 Solution | Winning solution for the GKI 2026 university division: next-byte modeling, habitat classification, time-series forecasting. | Python, PyTorch, XGBoost |
| Custom-Chatbot-Langchain | A free, custom LangChain chatbot build. | Python, LangChain |
- AI / ML: LLM fine-tuning (LoRA, SFT, DPO), RAG, quantization, vLLM, model deployment
- Security: bug bounty, mobile and binary reverse engineering, traffic interception, vulnerability research
- Tooling: Burp Suite, Frida, Ghidra, mitmproxy, Wireshark
Open to collaboration on applied AI and security research.