🧠 FlowOS

AI-powered cognitive optimization for neurodiverse minds.

Predict your focus. Enter flow on demand.

FlowOS is an AI-driven system concept designed to help people, especially individuals with ADHD, understand, predict, and optimize their cognitive performance. Unlike traditional productivity tools, FlowOS focuses on cognitive state modeling: transforming daily behavioral data into personalized focus insights and real-time interventions.

This repository currently contains the FlowOS pitch deck as a static HTML presentation.

🧑‍💻 Authors

Aria Shi GitHub, LinkedIn MSE @ Upenn | Machine Learning Engineer | BSc @ UCL | Physics & Math
Focus areas: Model Architecture, AI systems, Reinforcement learning, Applied ML infrastructure

Shufang Tan Researcher @ Children's Hospital of Philadelphia | PhD in Cognitive Science | MSc in Computer Science

Problem

People with ADHD and other neurodiverse conditions often struggle with:

• Inconsistent focus and energy levels
• Difficulty entering flow state
• Ineffective generic productivity tools
• Limited insight into how habits affect cognition

Existing tools track tasks, but they do not understand your brain.

Solution

FlowOS builds a personalized AI loop:

Behavior Data -> Pattern Learning -> Focus Prediction -> Actionable Interventions

Key capabilities:

• Behavior tracking: sleep, nutrition, hydration, activity, and mood
• Pattern discovery: identify personal focus triggers and blockers
• Focus prediction: estimate optimal cognitive windows throughout the day
• Flow Mode activation: provide real-time suggestions to help users enter deep focus

Core Insight

ADHD is not a discipline problem. It is a state regulation problem.

FlowOS adapts to the user's internal state instead of enforcing rigid routines.

System Architecture

Planned backend architecture:

app/
├── main.py                         # FastAPI entrypoint
├── schemas.py                      # Pydantic data models
├── routes/
│   ├── analyze.py                  # Behavior -> insights
│   ├── generate.py                 # Recommendations / Flow Mode
│   └── status.py                   # Job tracking
├── services/
│   ├── ingestion_service.py        # Data ingestion: manual + APIs
│   ├── feature_service.py          # Feature engineering
│   ├── modeling_service.py         # Focus prediction models
│   ├── recommendation_service.py   # Intervention logic
│   └── pipeline_service.py         # End-to-end orchestration
└── utils/
    ├── logger.py
    ├── metrics.py
    └── storage.py

Tech Stack

• Backend: FastAPI, Python
• ML/AI: PyTorch, Transformers, Scikit-learn
• Data: SQLite for MVP, scalable database later
• Infrastructure: Docker, async job handling
• APIs: OpenAI for analysis/reasoning, wearable integrations in future versions

Modeling Approach

Input signals:

• Sleep duration and quality
• Nutrition signals such as caffeine, sugar, and protein
• Physical activity
• Self-reported focus and mood

Methods:

• Time-series analysis
• Feature engineering with rolling averages and lag features
• Regression/classification models for focus score prediction
• Future reinforcement learning for intervention optimization

Core Pipeline

1. Collect daily behavioral data
2. Extract features and patterns
3. Predict focus score and focus windows
4. Generate recommendations
5. Collect feedback and improve the model

MVP Roadmap

v1: Current

• Manual logging
• Rule-based insights
• Basic focus estimation
• Static pitch deck

v2

• ML-based focus prediction
• Daily recommendations

v3

• Real-time adaptive system
• Flow Mode optimization
• Wearable integration

Use Cases

• ADHD students managing study sessions
• Engineers optimizing deep work time
• Creators improving consistency
• Anyone seeking high-performance cognitive optimization

Differentiation

Traditional Apps	FlowOS
Static schedules	Dynamic adaptation
Generic advice	Personalized insights
Task tracking	Cognitive state modeling
Reminders	Flow-state activation

Vision

Build the operating system for neurodiverse cognition.

Future directions:

• Real-time cognitive monitoring
• Cross-user pattern learning
• Integration with wearables such as Apple Watch and EEG devices
• B2B productivity and wellness solutions

Planned Metrics

• Focus score accuracy
• User retention and engagement
• Intervention success rate
• Recommendation pipeline latency, including p50 and p95

Why This Project

FlowOS combines:

• ML system design
• Real-world neurodiversity impact
• End-to-end production thinking

It reflects a shift from building models to building intelligent systems that adapt to humans.