Marie Humbert-Droz

ML Engineer & Healthcare AI Consultant

I build ML products for healthcare. I've done it twice as a founding engineer — both products shipped within a year.

Available for limited consulting with early-stage health tech teams.

Let's Talk

About

I help healthcare teams turn complex clinical data into intelligent, trustworthy ML systems — and I help founders figure out if their ML roadmap will actually work.

I'm currently Founding ML Engineer at Timeless Biotech, where I built our ovarian aging prediction platform (launching January 2026). Before that, I was Founding ML Scientist at Lighten, where I built an LLM-powered clinical abstraction platform that achieved 90%+ accuracy at scale.

I've walked the path from "interesting research idea" to "revenue-generating product" twice. I know where the pitfalls are, and I can help you avoid them.

I also publish technical tutorials on Towards AI, helping practitioners apply advanced ML to real-world healthcare problems.

Expertise

Core areas where I can help your team move faster, build smarter, and deliver real impact:

Large Language ModelsClinical NLPSurvival AnalysisTime-to-Event ModelingHealthcare Data SciencePredictive AnalyticsFemtech AI SolutionsStatistical ModelingDeep Learning0→1 Product Development

Featured Work

Timeless Biotech · 2025

Ovarian Aging Clock

Built a patent-pending ML platform from 0→1 that reduces time-to-menopause prediction uncertainty by nearly 10×. Developed the complete stack: core ML library, API, database, and frontend application.

→Built complete platform in 6 months — from research notebooks to production-ready full-stack application
→Patent pending, research papers in publication
→Reduces prediction uncertainty by nearly 10× using survival analysis and longitudinal cohort data
Read more
Lighten · 2024

Clinical LLM Abstraction Platform

Developed an LLM-powered system for clinical data abstraction from patient EHRs (often thousands of pages per patient). Leveraged prompt engineering and agentic workflows to extract complex clinical variables at scale.

→Achieved 90%+ F1 scores on complex clinical abstraction tasks
→Product generating revenue within first year
→Scalable, reliable abstraction for research and clinical workflows
2
Healthcare ML Products
Built from 0→1
<1 Yr
Avg Build Timeline
Concept to shipped product
90%+
F1 Score
Clinical abstraction at scale

How I Can Help

Let's TalkBook 15-min Intro

I help early-stage health tech founders de-risk their ML roadmap and accelerate time-to-market.

ML Strategy & De-risking Sprint

A focused 2-5 day engagement to validate your ML roadmap and identify risks before you invest heavily.

  • →Technical feasibility assessment
  • →ML roadmap validation & optimization
  • →Risk identification & mitigation strategies
  • →Written assessment with actionable recommendations

Ongoing ML Advisory

Strategic guidance and hands-on support as your ML product evolves (10-20 hours/month).

  • →Regular strategic ML guidance
  • →Architecture review & optimization
  • →Team mentorship & technical leadership
  • →Hiring support & due diligence assistance

Is this you?

✓You're a health tech founder with an ML vision but need validation
✓Your technical team needs a second opinion on feasibility
✓You're pre-Series A and can't afford a full-time ML hire yet
✓You need to de-risk your ML roadmap before raising your next round
Limited availability · Specialized in clinical AI and healthcare data science

Writing

I share practical lessons from building and evaluating real healthcare ML systems.

Oct 6, 2025

Can Local AI Keep Up with the Cloud? I Tested 8 Models on Clinical Data

Benchmarking 8 models for clinical extraction tasks.

Read on Towards AI
Sep 2, 2025

I Built a Local Clinical AI Agent from Scratch — Here’s How

Step-by-step guide to build a local discharge agent with GPT OSS.

Read on Towards AI
Jun 30, 2025

From Messy EHRs to 30-Day Readmission Predictions: Benchmarking 4 ML Models

Benchmarking macine learning models for a 30-day readmission task

Read on Towards AI
May 28, 2025

Detecting Hallucinations in Healthcare AI

Introducing 3 complementary methods to detect and mitigate hallucinations in AI

Read on Towards AI

Get in Touch

Looking to de-risk your ML roadmap or accelerate your health tech product? Let's talk.