Skip to main content

AI Engineer NYC 2026

​Join AI Engineer and Major League Hacking (MLH) on Sunday, October 11, for a full day of building, experimenting, and meeting fellow AI engineers.

Health-Maxxing (Health OS)

Snap a quick pic of what you're eating or drinking and in about 5 seconds get an estimate of plastic exposure and a safer alternative.

About the project

Plastic and chemical exposure can damage our health. Health-Maxxing is the first live module of my personal health operating system: snap a quick pic of what you’re eating or drinking, and in roughly 5 seconds you get an estimate of plastic exposure and a recommendation for safer alternatives. Easy, quick, seamless.

The problem. If you’re health-conscious, you make exposure decisions in seconds with no information. In my NYC hotel room I looked at the tap, felt uncertain about the quality, and took two plastic bottles from the front desk. No lookup, no comparison. Bottled water carries about 2x the microplastic particles of tap (Mason et al. 2018); running the cold tap for 30 seconds would have been the better call.

How it works

  1. See: Claude Opus 5.5 reads the photo into editable facts (brand, packaging, heated or not).
  2. Match: an estimate engine compares it with 618 lab samples from 236 products (PlasticList), falling back to similar products, then the category.
  3. Estimate: a relative tier with an uncertainty band and an evidence label (lab-measured, similar products, category, unknown), compared against reference levels from EFSA, EPA and NYC DEP.
  4. Swap / instructions: a lower-exposure option where the data backs one; for water, customised instructions for where you are (e.g. run the cold tap 30s+).

How we checked it: photo extraction eval on 71 labelled images (packaging read correctly on 17/17 product photos and 23/26 real-world snaps after fixing label errors); 21 automated tests; every cited figure links to its primary source.

Known limitations: estimates are relative, not safety verdicts (a “higher” food can still be under 1% of the EU tolerable daily intake); lab data is US-only with no NYC tap samples; short product names can produce false lab matches; user research so far is a self-interview. Next: our own lab studies, then bloodwork to see whether swaps change what’s in your body.

Credits: PlasticList, EFSA, US EPA, NYC DEP, Mason et al. 2018. Demo photos: JimmyStardust (CC BY-SA 4.0) and A1C Kylie Barrow (public domain), via Wikimedia Commons.

Built with

claude

AI tools

Coded withClaude Code

Prize Categories Entered

  • Event prize

    Best Overall

    Submitted

    AI-native and tackles a problem incredibly relevant and important for all of us: Health. And minimising exposure to forever chemicals that can damage us and our family.