Alexandra Abbas
Alexandra is a software engineer specialising in machine learning, with extensive experience in managing engineering teams. She recently served as a technical lead at Wise and has a strong background in data engineering and provisioning infrastructure for machine learning workloads.
Alexandra's research interests focus on understanding and mitigating harmful behaviours such as sycophancy and deception through novel techniques. She is also interested in expanding the open source toolkit for frontier model releases.
Alexandra is focusing on reducing sycophancy in frontier AI models by experimenting with techniques like fine tuning, activation steering and latent adversarial training.
