PhysioML – Adaptive Recovery Modeling Platform (Sleepmaxxing)

Local-first recovery intelligence built in C++

Sleepmaxxing is a local, explainable recovery analysis engine that learns from your personal health data over time.
It analyzes sleep, HRV, resting heart rate, stress, and activity to generate:

• Weekly recovery summaries
• A recovery score
• Short-term HRV predictions
• Actionable, human-readable guidance
• Model reliability metrics that improve as the system learns

No cloud. No subscriptions. No external user data.
Everything runs locally on your machine.

---

What This Project Does

Sleepmaxxing is designed to simulate how a real user would use a recovery product:

1. You log daily health data in a CSV
2. The system collects a baseline (first few weeks)
3. It begins making weekly predictions
4. Each prediction is evaluated against real outcomes
5. The model adapts and becomes more reliable over time

After ~5 evaluated weeks, you start seeing accuracy metrics instead of placeholders.

---

Demo

An 8-week end-to-end demo (compile → run → prediction → explanation) is included here:
https://www.loom.com/share/05b3960518ab4bc0ae42a1ede4591d17

---

Input Data Format

Sleepmaxxing reads a CSV file with one row per day.

CSV Columns

date,sleep_min,wake_min,hrv,rhr,stress,activity

---

Column Definitions

Column	Meaning
date	Calendar day of the record (YYYY-MM-DD)
sleep_min	Difference from your normal sleep duration (minutes)
wake_min	Difference from your normal wake time (minutes)
hrv	Heart Rate Variability (milliseconds)
rhr	Resting Heart Rate (beats per minute)
stress	Daily stress score (1 = low, 5 = high)
activity	Total steps for the day

---

Important

• sleep_min and wake_min are relative values, not raw timestamps
• This allows the system to reason about consistency and deviation, not absolute schedules.

Example Row

2025-01-01,10,-5,55,65,3,6500

###Interpretation

• Slept 10 minutes more than usual
• Woke up 5 minutes earlier than usual
• HRV = 55 ms
• RHR = 65 bpm
• Moderate stress
• 6,500 steps

---

How to Build & Run

1. Dependencies

C++ Dependencies

You must install:

• mlpack (C++)
• Armadillo (C++)

On macOS (Homebrew)

brew install mlpack armadillo

On Linux

sudo apt install libmlpack-dev libarmadillo-dev

---

2. Build the Project

From the project root:

make clean
make

This builds the sleepmaxxing binary.

3. Run Sleepmaxxing

./sleepmaxxing data/myhealth.csv YYYY-MM-DD

###Example

./sleepmaxxing data/myhealth.csv 2025-02-25

The date must exist in the CSV and represents the “current week” the system should analyze.

What You’ll See

A typical run outputs:

• Recovery Summary (Improving / Stable / Declining)
• Recovery Score (0–100)
• Confidence level
• Predicted HRV change (7-day horizon)
• Plain-English interpretation
• Primary physiological drivers
• Actionable guidance
• Model learning & reliability stats

---

Example

RECOVERY SCORE FOR THIS WEEK: 80
Strong Recovery

PREDICTED HRV DELTA (next 7 days): +1.11 ms

---

Model Reliability Explained

Sleepmaxxing is intentionally conservative with metrics.

• Predictions are evaluated only when real outcomes exist
• Accuracy metrics appear after enough evaluated predictions
• Early weeks prioritize transparency over false confidence

Typical progression

• Weeks 1–3 → Baseline collection
• Weeks 4–5 → Predictions + evaluation
• Week 5+ → Directional accuracy appears and improves

---

Features (V1)

• Local-only C++ recovery engine
• Weekly HRV prediction (ridge regression via mlpack)
• Recovery score (0–100)
• Driver attribution (sleep, HRV, RHR, stress)
• Confidence-aware explanations
• Incremental learning from past errors
• CSV-based data ingestion
• Clean Makefile-based build

---

Planned V2 Improvements

• Day-to-day predictions (not just weekly)

• Richer activity modeling

  • Currently: steps only
  • Planned: intensity, training load, recovery days

• More accurate ML models

  • Feature scaling
  • Better regularization
  • Per-user adaptation

• Improved recovery score calibration

• Graphs & visualization

  • Trends
  • Prediction vs outcome
  • HRV trajectories

Graph integration coming soon

Philosophy

Sleepmaxxing is built around a simple idea:

Personal health models should be local, explainable, and earned over time.

No black boxes.
No dashboards hiding uncertainty.
No cloud dependency.

Author

Built by Jordi Lopez
Early-stage systems + ML project focused on recovery intelligence.

If you’re interested in health tech, systems programming, or applied ML, feel free to reach out via my website:

https://jordisworld.me