Harper Ellis

Harper Ellis covers artificial intelligence for BitsFromBytes from San Francisco, where she spent four years as an NLP engineer at a mid-stage AI startup working on fine-tuning foundation models for legal and healthcare applications. She holds a master's in computer science from Stanford, contributes occasional corrections to the HuggingFace documentation, and maintains a small reading group for AI alignment papers that meets every two weeks at a Mission District coffee shop. Her writing for BitsFromBytes focuses on what large language models actually do versus what marketing copy says they do, which she thinks is the most under-covered topic in mainstream AI journalism. Harper is particularly interested in the gap between benchmark performance and real-world utility, and in the quiet ways model companies narrow the definition of safety over time. She is also a regular at weekly alignment meetups organized by various Bay Area research groups. Outside work she lives with two rescued cats and a bookshelf that her partner refuses to dust.

ChatGPT, Claude, Gemini, generative AI, prompt engineering, AI ethics, LLM research, alignment