Handshake AI Fellow
I created Physical Chemistry graduate-level problems to intentionally challenge advanced AI language models. I then devised detailed solutions, weighted rubrics, and hints aimed at helping these models learn from their reasoning errors. My work directly supported the AI models’ ability to solve complex, multi-step science questions. • Developed training prompts and rubrics tailored for AI reasoning evaluation. • Evaluated AI model responses for correctness and instruction-following. • Designed adversarial questions targeting specific weaknesses in AI reasoning. • Coordinated with AI research teams to optimize training setup.