YC Spring 2026 (P26) founder

Nathan H. Leung

Founder of Harbor, AI-native CRO for clinical trial operations

Harborhealthcare15 public signals

Company

Nathan H. Leung is listed as a founder of Harbor. Harbor is building an AI-native contract research organization (CRO) for clinical trials. We started with an AI-native system of record for clinical trial data (electronic data capture system), grew the software business to seven trials and ~$187k in contracted software revenue over the next twelve months, then expanded into full-service trial execution and signed our first CRO contract worth $1.93M over three years. Clinical trials consume 45% of all new drug and device development spend. While some of that goes to software, sites, and patients, the largest line items are services: project management, monitoring, data management, site coordination, source verification, query resolution, record reconciliation, and analysis. Today, drug and device companies outsource clinical operations to legacy CROs and their massive teams, which is why the CRO market is $82B annually. Harbor delivers the regulated services companies buy from legacy CROs, but built around our AI-native system of record from day one. Our software already automates the core data workflows: study build, data capture, query generation, monitoring, signing, and database lock. Because we own the system of record, we can expand into the surrounding service workflows: site contracting and billing, subject engagement and retention, statistics, reporting, and more. We built Harbor because we lived through the problem. Albert spent four years in clinical trials and regulatory strategy at a medical device startup, taking an unproven wearable from first-in-human trials to FDA authorization. Nate brings the technical depth, having built scalable software at Google and Ramp and served as employee #1 at a previous YC startup.

Founder traction evidence

Public posts and activity attributed directly to this founder.

  1. X

    Excited to announce I am no longer unemployed! Last June, I quit my job, emptied my apartment in San Francisco, and drove eight hours south to the suburbs of San Diego, where I moved into my friend Albert’s apartment. For nine months, we lived on Tasty Bite Madras lentils and Show more 0:03 / 2:00

    Excited to announce I am no longer unemployed! Last June, I quit my job, emptied my apartment in San Francisco, and drove eight hours south to the suburbs of San Diego, where I moved into my friend Albert’s apartment. For nine months, we lived on Tasty Bite Madras lentils and...

  2. X

    Introducing slop-loader: a prompt-to-code loader for Next.js/Webpack. Import LLM outputs as build-time content, storing only raw prompts in your repo. Just `import Header from "./Header__prompt.tsx"` and render `<Header />`. https:// github.com/nathanhleung/s lop-loader … s/o @finerflame for idea

    Introducing slop-loader: a prompt-to-code loader for Next.js/Webpack. Import LLM outputs as build-time content, storing only raw prompts in your repo. Just `import Header from "./Header__prompt.tsx"` and render `<Header />`. https:// github.com/nathanhleung/s lop-loader … s/o...

  3. X

    10y ago today, my high school principal suspended me for making an app that let my classmates see who was in their classes before school started Every day I think about what life would've been like if I'd turned it into a GPT wrapper and raised 5 at 50 She still mogs me on base

    10y ago today, my high school principal suspended me for making an app that let my classmates see who was in their classes before school started Every day I think about what life would've been like if I'd turned it into a GPT wrapper and raised 5 at 50 She still mogs me on base

  4. X

    Harbor automatically extracts data from source document scans, so clinical research teams can spend more time on the human parts of a clinical trial. GIF

    Harbor automatically extracts data from source document scans, so clinical research teams can spend more time on the human parts of a clinical trial. GIF

  5. X

    6 weeks into YC P26 with @runharbor . Here's what’s happening on the ground: • Cursor seems a little cooked... VSCode is back. • Codex > Claude Code (at least until OpenAI clamps down on usage limits) • The "fail upward" path to AI labs feels like it has lowered startup risk Show more

    6 weeks into YC P26 with @runharbor . Here's what’s happening on the ground: • Cursor seems a little cooked... VSCode is back. • Codex > Claude Code (at least until OpenAI clamps down on usage limits) • The "fail upward" path to AI labs feels like it has lowered startup risk...

  6. X

    waterloo interns in san francisco: do you ever wonder what you left behind? 0:04 / 4:35

    waterloo interns in san francisco: do you ever wonder what you left behind? 0:04 / 4:35

  7. X

    Replying to @nathanhleung Link your trial protocol and Harbor analyzes and compiles a fully compliant study database complete with study events, forms, and fields. GIF

    Replying to @nathanhleung Link your trial protocol and Harbor analyzes and compiles a fully compliant study database complete with study events, forms, and fields. GIF

  8. X

    For monitors, Harbor shows database values and source documents in the same platform, with confidence scores highlighting potential issues. Instead of spending the afternoon comparing values from a paper binder, monitors can spend their time in the clinic keeping things running Show more GIF

    For monitors, Harbor shows database values and source documents in the same platform, with confidence scores highlighting potential issues. Instead of spending the afternoon comparing values from a paper binder, monitors can spend their time in the clinic keeping things running...