Hi 👋,
I’m a Member of Technical Staff at the UK AI Security Institute, where I work on frontier-model evaluation and research engineering. My work spans AI safety research, model training and evaluation, distributed infrastructure, and early-stage AI products. Previously, I built production systems at Confluent and co-founded an applied-AI company through Entrepreneurs First.
I’m particularly interested in technically difficult problems that require moving between research, engineering, and product development.
Phantom Transfer: Data Poisoning can Survive Data-Level Defences
Co-first author · 2026
Paper · Code · Alignment Forum · LessWrong
Quantifying Frontier LLM Capabilities for Container Sandbox Escape
ICML 2026 Spotlight · 2026
Paper · Code · Blogpost
Weak Constraint Gaussian Processes for Optimal Sensor Placement
Journal of Computational Science · 2020
Paper
Data Learning: Integrating Data Assimilation and Machine Learning
Journal of Computational Science · 2022
Paper
Multimodal Property Search — natural-language search over images and listing text
Gaussian-Process Sensor Placement — MSc research project
Whisper & Llama Fine-tuning — domain adaptation experiments
Two-Tower Document Retrieval — semantic retrieval system
Robot Brain Cluster — distributed compute / robotics project
Budgeting App — UX research and mobile product design
Camping Store — responsive e-commerce design
Examenstein — earlier product concept
Image Captioning Generator — deep-learning residency project
Connect with me over LinkedIn or email me at durtolga1@gmail.com