Vicki Boykis
Also known as Normcore Tech
Vicki Boykis builds machine learning systems and explains their engineering through public essays, projects, and conference talks. Her earlier work includes Comcast, Automattic, Tumblr, Duo, and Mozilla.ai. A 2022 essay focused on Git, SQL, and command line tools as foundations for data work. She also helped run Normconf. Her detailed explanation What Are Embeddings? and the Viberary book recommendation project connect model representations with practical information retrieval. In a May 2024 PyCon Italia talk, she discussed LLM evaluation and infrastructure at Mozilla.ai. She is a founding engineer in recommendation and personalisation. Her essays link models with data pipelines, search, recommendation, and the routine work needed to operate a useful system.
Timeline
-
Earlier data and ML work
Building systems across companies
Boykis has worked at Comcast, Automattic, Tumblr, Duo and Mozilla.ai. Her 2022 tools essay describes writing data-centred code across varied organisations and languages.
-
January 2022
Teaching durable engineering tools
In Git, SQL, CLI, Boykis discussed three tools she considered essential for data work. She drew on experience with different languages, databases, and operating constraints. The article placed routine engineering skills within the work of building data systems, giving context for her later AI explanations.
-
2022
Bringing practitioners together
She ran the ML and data conference Normconf and gave its 2022 keynote.
-
2023
Explaining embeddings
Boykis published What Are Embeddings?, a detailed account of numerical representations used in search and recommendation. It connected the idea with practical systems and model families. Her public Viberary project explored recommending books by a requested mood, providing an application alongside the technical explanation.
-
May 2024
Connecting LLMs with older engineering problems
Her PyCon Italia talk, published as a near-transcript, described work at Mozilla.ai on model evaluation and selection. She linked language model infrastructure with earlier ML systems in security and recommendations. The discussion gave builders a practical frame for errors, iteration, and operational concerns.
-
Current public work
Personalisation and information retrieval
Boykis is a founding ML engineer focused on recommendation and retrieval. She explains how these systems are assembled and operated, connecting model behaviour with the surrounding engineering work.
Sources and credits
Photo: Vicki Boykis / personal website. License: Not stated by source. Source: vickiboykis.com. Cropped to a square and resized. Ask us to remove or replace this photo.