I'm Luke Harries
I lead growth at ElevenLabs, one of the fastest growing startups in the world. I do a mix of marketing, product, engineering and developer experience.
Alongside, I angel invest in startups. I'm proud to have backed some exceptional companies including Lovable and Runna.
Previously, I was interim head of product at PostHog and co-founded Fella Health, a YC backed HealthTech company that treats thousands of patients per month. I started my career at Microsoft Research working on reinforcement learning.
I studied pre-med at Cambridge University. My parents are still hopeful that I might become a doctor.
Your first growth hire
·
Who should they be? How should you evaluate them? Do you go for a generalist or a specialist? And how should you think about agencies, contractors and advisors?
Read more →Launch and launch again — how to launch your products
·
A launch is a high-leverage moment to tell the world about your new product or feature. Here's how to make it count.
Read more →SEO for startups — a guide for founders and engineers
·
The guide I wish I had when learning about SEO. We cover understanding the user's intent, solving the user's intent, and getting your pages ranking.
Read more →10x engineers talk to users
·
Engineers should speak directly to users. Having a PM separating engineers from users is an anti-pattern.
Read more →GPT Automator, LangChain and Tool Engineering - EA London Hackathon
·
Building a voice-controlled AI assistant that can control your computer — what could go wrong?
Read more →YC - My experience and advice for the application and interview
·
If you are an early-stage company, I think YC has incredible value and should still be the default for most founders.
Read more →Building credit - why it matters and how to do it? (UK)
·
Learn how to build your credit score from scratch and avoid the catch-22 of needing credit to get credit.
Read more →Launching a HIPAA compliant MVP
·
This is a guide for early-stage HealthTech startups who want to launch an MVP quickly and cheaply while following HIPAA regulation. It's the guide that I wish I had when setting up infrastructure for COVID-19 drive-through testing.
Read more →What's the God mode?
·
Working backwards from the theoretical optimum to unlock better product design.
Read more →Useful resources for learning programming and machine learning
·
A collection of tools, guides, and courses that I found useful or wish I had found earlier.
Read more →The hidden power of HackerNews search
·
Hacker News is a treasure trove of information. Here's how to use it properly.
Read more →
WhatsApp MCP Server
·
A Model Context Protocol (MCP) server to connect LLMs to your personal WhatsApp account.
You can send and read your personal Whatsapp messages (including images, videos, documents, and audio messages), search your contacts and send messages.
I built it in a weekend and it went surprisingly viral. It's now one of the most used MCP servers and has 3k+ stars on GitHub. Code and instructions are from here.
VideoToSoundEffects.com
·
For the launch of the ElevenLabs Sound Effects API, Austin Malerba and I built VideoToSoundEffects.com. It allows you to upload any video and generate a sound effect for it. We built it using ElevenLabs, OpenAI's GPT-4o, and WASM.
We cranked it out in 1 day. The launch video was seen by 100k+ people, 25k+ people visited the site, and it was feature in Tom's guide.
You can see the demo video here and the code here.
Computer control with OpenAI's GPT-3
·
Chidi Williams and I built a tool to physically control your computer through your voice using Whisper, LangChain, and GPT3. It can open and interact with applications, click around in chrome, and synthesize information. You can see the video and read more here.
Reinforcement learning for functional software testing
·
We built a reinforcement learning system for testing software, specifically the Windows operating system. It combined Graph Neural Networks with Reinforcement Learning to trigger desired software functionality for testing. I co-led the development while at Microsoft Research. The paper was accepted as a workshop paper at ICML 2019 and is available here.
3D environment for evaluating AI agent generalization
·
MazeExplorer is a customisable 3D benchmark for assessing generalisation in Reinforcement Learning. It is based on the 3D first-person game Doom and the open-source environment VizDoom. I co-led the development while at Microsoft Research.
We presented MazeExplorer as a poster presentation at NeurIPS 2023 and it was published in the IEEE Conference on Games. The code is on GitHub.
On-device pose classification for physiotherapy
·
We developed a tool to gamify physiotherapy workouts by using machine learning to classify poses. The aim is to increase adherence of at home workouts. All training and prediction is done in-browser using Tensorflow.js, reducing any reliance on servers and meaning your images never leave the computer - maintaining privacy. Winner of HackCambridge 2019.
Deep learning for detecting cancer mutations
·
I completed my master's thesis with Cambridge Cancer Genomics (YC S17), leading the development of SomaticNet - a novel method for detecting cancer mutations. The resulting paper was accepted at the Machine Learning for Health workshop at NeurIPS 2018.
AI caption generation for visually impaired people through a chrome extension
·
For the millions of people with Visual Impairments, images online are inaccessible. We created a chrome extension which uses machine learning to automatically generate the missing image captions. The captions describe what's in the photo, who they are, and what any text says. Winner of HackCambridge's Microsoft Prize 2018.
Google Cardboard to screen for Macular Degeneration (proof-of-concept)
·
We created a proof of concept for faster and cheaper screening of Macular Degeneration. It is built for Google Cardboard which turns any phone into a VR headset.
The solution has two parts: mapping the functional visual field; and simulating what it's like living with MD by warping a real-time camera feed. Winner of Allia's Serious Impact Hackathon 2017. The code is on GitHub.
Platform for university sports
·
During my final year at Cambridge we built Proteam. The iOS and Android apps allowed students to follow teams, get notified of upcoming games, create team-sheets and much more.
We grew the platform to 9,000+ users across 6 Universities.