FitSenseAI

AI-powered fitness coaching application focused on personalized workouts, progress tracking, and health-aware guidance.

What’s In This Repository

This repository currently includes:

• project planning artifacts,
• database schema design for the FitSenseAI domain,
• an implemented synthetic-data MLOps pipeline in Data-Pipeline/ (data generation -> teacher LLM -> distillation dataset -> validation/monitoring),
• Airflow DAG orchestration for the pipeline.

Project Scope

FitSenseAI is designed to:

• Create and adapt workout plans based on user goals, conditions, and workout history.
• Log workout execution (exercises, sets, reps, weight, RIR, notes).
• Store relevant user health context (medical profile, medications, allergies, injuries).
• Recommend daily maintenance calories and allow optional daily calorie intake logging.
• Recommend target sleep duration and allow optional daily sleep-duration logging.
• Allow users to log body weight any time for progress tracking.

Out of scope for current data model:

• Detailed nutrition macro tracking (protein/carbs/fats).
• Hydration tracking.

Repository Structure

FitSenseAI/
  README.md
  FitSense_AI_Project_Plan.md
  FitSense_AI_Project_Scoping_Complete-1.pdf
  MLOPS-1-2_FitSenseAI_Execution_Guide.md
  Data-Pipeline/
    dags/
    scripts/
    tests/
    data/
    logs/
    params.yaml
    requirements.txt
    dvc.yaml
  database/
    database_design.dbml
    postgresql.sql
    mysql.sql
    UML_diagram.png

Database Design

Primary schema file:

• database/database_design.dbml

Live ER diagram (no upload needed):

• https://dbdiagram.io/d/FitSenseAI-69850002bd82f5fce2cfe02c

Core model areas:

• User and goals: users, goals, user_goals
• Health context: conditions, user_conditions, user_profiles, user_medical_profiles, user_medications, user_allergies
• Workouts: workout_plans, plan_exercises, plan_sets, workouts, workout_exercises, workout_sets
• Guidance + tracking:
  • Calories: calorie_targets, calorie_intake_logs
  • Sleep: sleep_targets, sleep_duration_logs
  • Weight: weight_logs
• AI interactions: ai_interactions

Quick Start (Schema)

1. Open the live diagram: UML Diagram
2. No DBML upload is required for viewers; they can access the schema directly from the link.
3. Use database/postgresql.sql (PostgreSQL) or database/mysql.sql (MySQL) as the base SQL export for database setup.
4. Prefer PostgreSQL for implementation (recommended for this project’s relational complexity and future analytics needs).

Planned Architecture (High Level)

• Backend API: user onboarding, goal capture, workout planning, logging, and AI endpoints.
• Mobile/Web client: plan viewing, workout execution logging, and daily check-ins (calories/sleep/weight).
• AI layer: plan generation/adaptation and conversational guidance.
• Data layer: relational DB for user/workout/health data and model interaction logs.

See FitSense_AI_Project_Plan.md for phase-wise execution details.

Architecture Diagram (Detailed)

For a more complete write-up, see docs/Architecture.md.

Data Pipeline (Overview)

FitSenseAI includes an end-to-end synthetic-data MLOps pipeline under Data-Pipeline/ that covers:

• synthetic profile/workout/health data generation,
• synthetic query generation for a teacher LLM,
• teacher response capture and storage,
• distillation dataset creation (train/val/test JSONL),
• validation, statistics, and anomaly detection,
• Airflow DAG orchestration for the full workflow.

Primary docs:

• Data-Pipeline/README.md for the full pipeline usage and Airflow commands

Pipeline Component Diagram

Airflow DAG Diagram

Pipeline Quick Start

python -m venv .venv
source .venv/bin/activate
pip install -r Data-Pipeline/requirements.txt

# bootstrap
python Data-Pipeline/scripts/bootstrap_phase1.py

# run pipeline stages (script-by-script)
python Data-Pipeline/scripts/generate_synthetic_profiles.py
python Data-Pipeline/scripts/generate_synthetic_workouts.py
python Data-Pipeline/scripts/generate_synthetic_queries.py
python Data-Pipeline/scripts/call_teacher_llm.py
python Data-Pipeline/scripts/build_distillation_dataset.py
python Data-Pipeline/scripts/validate_data.py
python Data-Pipeline/scripts/compute_stats.py
python Data-Pipeline/scripts/detect_anomalies.py

Bootstrap outputs:

• Data-Pipeline/data/reports/phase1_bootstrap.json
• Data-Pipeline/logs/pipeline.log

Roadmap Snapshot

• Phase 1: Foundation and data schemas
• Phase 2: Model development (teacher/student workflow)
• Phase 3: Backend + app MVP
• Phase 4: Adaptation engine and instrumentation
• Phase 5: Safety, monitoring, hardening
• Phase 6: Pilot, iteration, final validation