Commoditizing health

Commoditizing health

Colon Cancer

Due to my goal of improving the performance of some biotech algorithms 1000x, I have come to the conclusion that we live in the stone age for health. I should not have to do anything to be healthy.

We need a zero-friction, 24x7x365, full-body, real-time, atomic-level precision, health diagnostic tool that knows when something is messing us up, like a hemorrhoid that might be cancer, and fixes it- all without me having to think about it.

And just like how Amazon + SpaceX bootstrapped knowledge in a new domain through business + engineering, not science, we need a flywheel that does this. A self-sustaining, network-growing data flywheel that both creates the mathematical equations while getting live feedback.

Reddit threads that inspired me:

https://www.reddit.com/r/cancer/comments/90vruy/serious_redditors_with_colon_cancer_what_were/

https://www.reddit.com/r/ostomy/comments/18kucqu/people_who_got_colon_cancer_in_their_20s_or_30s/

image
image
image

The Device

1. Signal Acquisition

What can we sense cheaply, non-invasively, and continuously- today?

  • OpenBCI: EEG/EMG/ECG open source biosensors
  • Empatica E4 SDK: Physiological data streams (heart rate, EDA, temp)
  • HealthyPi v5: Raspberry Pi-based open-source vital signs monitor
  • OpenHealth: Wearable platform for real-time biomedical data

2. Data Unification + Inference

A NN that unifies scattered biological signals and discovers patterns.

  • OpenMined / PySyft: Privacy-preserving federated learning
  • Nilearn + MONAI: Medical imaging + signal processing
  • NightShift / Circadia: Circadian + chronobiology signal modeling
  • VitalDB: ICU signal dataset for training real-time predictive models

3. Disease Prediction + Interpretation

Focused pipelines we can open-source, scale, and later integrate.

  • DeepHealth (MIT) or DeepMed: Cancer classification pipelines
  • lifelines: Survival models for predicting disease progression
  • GenoML / BioAutoML: Genomic machine learning made modular
  • FHE-ML (Susakshai-style): Secure inference without exposing raw data

4. Real-Time Infrastructure

After we have captured all information, we need to process signals and actually garner useful patterns to build systems off of.

  • Home Assistant + MQTT + Node-RED: Modular edge monitoring
  • Edge Impulse: Real-time ML inference at the edge (on-device health AI)
  • Prometheus + Grafana: Real-time observability stack for your own body

Starter Project:

A pipeline that:

  • Takes in EKG, HRV, SpO2 from OpenBCI or HealthyPi
  • Runs real-time models on edge (via Edge Impulse or PyTorch Mobile)
  • Streams output to a Grafana dashboard with alerts (e.g., abnormal arrhythmia)
  • Adds federated learning support via OpenMined to protect user privacy
  • Long-term: plug in blood biomarkers, microbiome data, genome