I’m an ML engineer with four years of experience, now transitioning into full-time AI safety. I left my job and am self-funding this transition. I am focussing on concrete actions and deliverables that can make AI safer at internal and external levels through improved behavioural analysis, evaluation and systemic safeguards that also account for alternative pathways, architecture and future advanced capabilities. For me, the most effective approach is a combination of things that work best rather than a single technique, and it should also account for domain shifting. I want to connect the gap  between AI Safety research and result delivery into action. 

I completed BlueDot’s Technical AI Safety course and have three research projects going on right no related to LLM collusion, AI judge reliability and how AI systems respond to one another.