Make your LLM app a Domain Expert: How to Build an LLM-Native Expert System
A talk I’ve been invited to give at the AI Engineer Summit on June 3rd 2025 - I’ll update the content here shortly.
Chris Lovejoy
About Me
I'm part of the founding team at Anterior, a Sequoia/NEA-backed start-up building an AI clinical brain. I've been a medical doctor, academic researcher, entrepreneur and AI/ML engineer.
Popular
- Machine Learning for Healthcare video series (50k views)
- The Academic Knowledge Management (AKM) system that 10x'd my research productivity
- The Truth Is In The Title? Video Title Generation as a novel training objective for video summarisation
- AI versus clinicians: systematic review of design, reporting standards, and claims of deep learning studies
- I created my own YouTube algorithm (to stop me wasting time) (100k+ views)
- How I became a Data Scientist during COVID (as a Doctor): 5 takeaways from an unconventional career transition
Recent Essays
Who to hire first for your LLM product team? 3 principles for building LLM product teams
Team is the single biggest predictor of success – but there’s no ‘playbook’ yet for what teams building LLM products should look like. At Anterior, we’ve gone from first mover in the health insura...
My winding, unplanned path from Doctor to AI Startups (and all the steps in between)
In 2020, as the peak from the first wave of COVID subsided in the UK, I worked my final shift as a medical doctor. Since then, I’ve worked as a data scientist, founded a VC-backed start-up, been an...
Mission-Critical Evals at Scale: Learnings from 100,000 medical decisions
Here’s a talk I gave for the AI Engineer Summit (22nd Feb 2025). Recording Summary Slides
Relying only on offline evals is playing with fire: why you must evaluate on live production data
So you’ve followed the rule book: you’ve created your labelled datasets, used them to iterate against as you improve your LLM product, and you’re going live with early customers. Metrics look good,...
I fired my personal coach and built one with AI - it knows me better, gives great advice and saves me $$$ every month
Last year, I tested out building an AI performance coach to replace my human one. It’s much better so I cancelled my human one altogether. It knows me better, gives great advice and improves itself...
5 ways to reduce the risk and impact of hallucinations in your LLM product
I’ve been building apps powered by LLMs for the past 2 years - mostly in healthcare where the stakes are high. The single most common question I get is: “So.. how do you deal with hallucinatio...
Learning AI - where to start in 2024
An academic medical doctor recently asked me: “I want to learn about AI - should I learn to code? should I learn maths?” My response was: It depends what type of ML you want to do. I find it help...
Can LLMs make idea-sharing truly frictionless?
I want to share valuable ideas on the internet but I don’t want that to be my main thing. I’d rather it be a byproduct of cool stuff that I’m doing. The problem is that if I’m heads-down building,...
What makes a good meeting? A high rate of info exchange and/or decisions
In the Art and Business of Online Writing, Nicolas Cole describes the “rate of revelation” as a key metric for how engaging written work is: the rate at which the reader discovers new information. ...