About me

I'm Seraphina Goldfarb-Tarrant, and I -

• am the Head of Safety at Cohere, where I try my best to evaluate and mitigate any societal harms cause by language models, and to make LMs more trustworthy,
• have a PhD in Informatics (CS) from the University of Edinburgh in ILCC, advised by Adam Lopez. I focused on Fairness in NLP, my thesis is: Fairness in Transfer Learning for NLP
• have an MSc in NLP from University of Washington
• did research in Creative Neural Natural Language Generation (a mouthful, aka NLG for short), at the Information Sciences Institute in the lab of Nanyun (Violet) Peng.
• think this bio reminds me a bit of this.

I am currently looking for interns (I usually am, but also currently!). If you are interested, please email me a few paragraphs up to 2 pages on a research idea or proposal, and let's chat.

I have a varied background: I live in the UK and work in AI now, but I have a BA in Ancient Greek, and I have lived in many places around the world and worked in a few different fields. My homes have been Tokyo, NYC, San Francisco, Seattle, Shanghai - in those places I've worked both for Google and on smaller projects.

As well as being extremely passionate about NLP, I am a professional sailor and spent a few years sailing the North Sea. I also study Rinzai Zen, which is the type with the riddles (koans) and samurai associations.

Feel free to check out my CV and my research/publications. Because I am not good at updating this site, my list of publications may be more up to date on my Google Scholar.

I started a blog in my MSc, but as is the case with all blogs, I wrote one article ever and then abandoned the enterprise. I do write some content at Cohere still. For instance, I wrote the 7 things to Know about Safety in NLP, which marketing rebranded as the Enterprise Guide to AI Safety, with no real change in content.

I am quite passionate about explaining things well, and frequently give talks and lectures. I mentor students (and sometimes co-supervise) and I accept interns at Cohere. If any of these things interest you, please feel free to get in touch!

If you need a blurb for an event I am speaking at, you may nick this one below:
Dr. Seraphina Goldfarb-Tarrant is the Head of Safety at Cohere, where she works on both the practice and the theory of evaluating and mitigating harms from LLMs. She has a PhD in NLP from the University of Edinburgh, with a focus on Fairness, and an MSc from UW, with a focus on generation. Her research interests include the intersection of fairness with robustness and generalisation, cross-lingual transfer, and causal analysis. Before her PhD, she worked as a Research Engineer at the Information Sciences Institute, and for many years at Google in Tokyo, NYC, and Shanghai. She also spent two years as a sailor in the North Sea.