Chris Lovejoy Portfolio
Member of Technical Staff at Anthropic
Former UK medical doctor, now Member of Technical Staff at Anthropic after the founding team at Anterior, a healthcare LLM startup. The homepage leads with a one-line thesis (AI products for complex, domain-specific verticals), then a "Popular" list where every item carries its own proof: 100k+ views, 50k+ views, 1,000+ citations.
Location
Somewhere on planet Earth, most likely near a laptop.
Companies
Credentials
- University of Cambridge — Medicine (2011–2017)
Skills
Notable
- Co-author of "Artificial intelligence versus clinicians", a systematic review in the BMJ (2020), which his site lists at 1,000+ citations
- Speaker at the AI Engineer Summit (Feb 2025) and the AI Engineer World's Fair (June 2025 and July 2026)
- YouTube Video Finder: 525 GitHub stars, and 343 points on Hacker News for the write-up
- Author of "The Modern Medical Student Manual", written while a junior doctor
What makes it work
A breakdown of the choices that make this portfolio stand out.
A niche stated in the first sentence
The About line says exactly what he builds: AI products for complex, domain-specific verticals like healthcare and life sciences. Every talk and essay below it (domain-native LLM apps, evals across 100,000 medical decisions, leveraging domain experts) reinforces that one position instead of a generic "AI engineer" skills list.
Proof attached to each headline item
The "Popular" list pairs each piece with a number: 100k+ views on the vertical-AI playbook talk, 50k+ on the ML for Healthcare video series, 1,000+ citations on a BMJ systematic review he co-authored. The BMJ authorship checks out on PubMed, and the YouTube-ranking side project has 525 GitHub stars.
A career change told openly, with dates
A dated essay walks through the change: roughly three years from first doubting a medical career to a first data-science role, a final shift as a doctor in 2020, then a VC-backed startup, consulting for 10+ health AI companies, and Anterior. The medical background becomes the reason his vertical-AI work is credible, not a gap to explain away.
What AI Engineers can take from this
Specific, actionable tips to apply to your own portfolio — no generic advice.
State your niche in one sentence at the top of the page, then make every featured piece of work support it.
Put a verifiable number next to each headline project (views, citations, stars) so a reader can see why it's featured without clicking through.
If you changed careers, write the transition up with dates and treat your previous field as domain expertise, not a detour.
Projects
A doctor who left medicine in 2020 and now builds vertical AI at Anthropic. The projects below run from a side project that ranks YouTube videos to peer-reviewed research, and each one is backed by a public repo or paper.
YouTube Video Finder
A personal replacement for the YouTube recommendation feed. A Python script queries the YouTube Data API for video- and channel-level stats, scores each video with his own ranking formula, and an AWS Lambda job automatically emails him the top-ranked results so he can pick a week's viewing in one batch instead of scrolling.
The ranking reasoning is written out: raw view count gets outweighed by views relative to channel size, and clickbait is treated as a failure mode. Readers see how he decides what to rank, not just the code that does it.
Scores videos on view count relative to the channel's subscriber count, not raw popularity
An AWS Lambda job delivers the ranked picks by email
Full write-up covers the ranking design, not just the code
AI versus clinicians: systematic review (BMJ)
A systematic review of studies comparing diagnostic deep learning algorithms for medical imaging with expert clinicians. It covers design, reporting standards, risk of bias and the claims those studies make, drawing on Medline, Embase, the Cochrane trials register and the WHO trial registry from 2010 to June 2019.
Evaluating whether an AI model's claims hold up is the same skill production AI evals need. This is that skill applied to published medical AI research and peer-reviewed.
Peer-reviewed in the BMJ alongside co-authors including Eric Topol and John Ioannidis
Audits the claims of AI-versus-clinician studies, not just their headline accuracy
Covers trial registrations as well as published studies
Video Title Generation and the YouTube Titles and Transcripts dataset
Proposes video title generation as a new task and releases the YouTube Titles and Transcripts (YTT) dataset: 17,886 video titles with their transcripts. The paper treats title generation both as forward prediction and as summarisation, and compares ways of pre-processing transcripts to pull out the relevant content.
He defined a new task, built the dataset for it, and evaluated the outputs with human raters. That is the full research loop, not a fine-tuning exercise on someone else's benchmark.
Introduces a new NLP task and its own 17,886-example dataset
Compares forward-predictive and summarisation framings
Human evaluation rather than automatic metrics alone
More AI Engineer portfolios
Ready to build your portfolio?
Follow the AI Engineer roadmap — skills, projects, and timeline to get hired.