A focused AI health-product exploration for understanding how training, fueling, recovery, and body-weight trends influence strength progression. I am building it to learn how AI, product analytics, health data, and frontend development can be used toward a real personal health goal:
Understanding what helps me train better, recover better, and make more informed strength-training decisions.
Explore: Updated Overview
This is not presented as a finished medical product, a validated coaching system, or a replacement for professional guidance. It is a transparent experiment in applying AI and data analysis to a problem I genuinely care about.
Current status: Product concept and interactive UI prototype. The measurement system, data architecture, recommendation logic, and AI guardrails are being developed and tested.
Strength Intelligence is organized as a living product case study. Each document explores a different part of the product, research, analytics, and system design.
| Document | Description |
|---|---|
| Case Study | Overview of the problem, approach, current progress, limitations, and lessons learned. |
| Product Vision | Long-term product direction, intended user experience, and future capabilities. |
| Product Requirements | Product goals, user needs, requirements, success metrics, and scope. |
| Measurement Framework | Primary outcomes, supporting metrics, guardrails, and measurement logic. |
| Data Model | Proposed entities, relationships, schemas, and data definitions. |
| Analytics | Planned SQL, Python, longitudinal analyses, and experiment concepts. |
| AI Framework | Recommendation logic, context design, uncertainty handling, and AI evaluation. |
| Research | Supporting evidence, research questions, assumptions, and limitations. |
| Design | Interface direction, information architecture, and product interaction decisions. |
| System | System design, system architecture, and data flow. |
I have trained consistently for years and already collect useful information across different tools.
My current data comes from:
- Apple Health, which includes sleep, steps, heart rate, activity, body weight, and nutrition data imported from connected apps
- My workout journal, where I would record exercises, sets, reps, weight, effort, and session notes (currently not updated nor available)
The problem is that these sources remain disconnected.
I may know that I slept less, ate fewer carbohydrates, walked more than usual, and had a weaker workout. But I still have to decide:
- Which factors actually mattered?
- Is this a real pattern or a one-time event?
- Should I progress, repeat, or adjust the next session?
- What should I test next?
My journal tells me what happened in the gym. Apple Health gives me context about sleep, activity, recovery, and nutrition. I still have to manually decide whether those signals affected my performance.
Strength Intelligence explores whether these data sources can be combined into something more useful.
I am personally interested in the intersection of:
- Strength training
- Human performance
- Health technology
- Product development
- Data analysis
- AI-assisted decision-making
This project is grounded in my own routine and questions.
I want to understand things such as:
- Why are some workouts noticeably stronger than others?
- Is my sleep affecting specific lifts?
- Is a calorie deficit limiting progression?
- When should I increase weight?
- When should I repeat a session?
- Which recovery and nutrition patterns are actually useful for me?
Because the project uses a real problem from my own life, it gives me a practical environment for learning product analytics, research, data science, systems design, frontend development, and AI product development.
The goal is not to create another workout tracker.
The goal is to explore how disconnected health and workout data can be turned into:
- Clear measurements
- Understandable insights
- Transparent recommendations
- Better questions for future research
Phase 1:
- Problem framing
- Data audit
- Product requirements
- Prototype UI
- Measurement framework
Phase 2:
- Workout logging system
- Structured Notion database
- Apple Health import
- Initial analytics
Phase 3:
- Longitudinal analysis
- Insight validation
- Recommendation engine
Phase 4:
- User testing
- Additional data sources
- AI evaluation
flowchart LR
A[Apple Health] --> C[Unified Data Model]
B[Workout Journal] --> C
C --> D[Analytics Layer]
D --> E[Product Insights]
D --> F[Recommendation Engine]
E --> G[Dashboard]
F --> G
| Area | What is shown |
|---|---|
| Product | Problem framing, product requirements, roadmap |
| Analytics | Metrics, KPIs, SQL, dashboards |
| Data Science | Exploration, prediction, time-series thinking |
| Research | User questions, evidence review, limitations |
| Systems | Data flow, architecture, interfaces |
| AI | Context building and recommendation logic |
| Frontend | Visual product experience and component planning |
- Apple Health for sleep, activity, body weight, heart rate, nutrition, and recovery context.
- Workout Journal for exercises, sets, reps, weight, effort, and session notes.
- Overview
- Progressive Overload
- Fueling
- Session Analysis
- Insights
Imagine that a lower-body workout performs below its recent baseline.
A normal workout tracker may only show the completed exercises, sets, reps, and weight.
Strength Intelligence adds context:
flowchart LR
A[Shorter Sleep] --> E[Session Context]
B[Higher Daily Activity] --> E
C[Lower Carbohydrate Intake] --> E
D[Recent Training Load] --> E
E --> F[Below-Baseline Performance]
F --> G[Repeat Weight and Test Fueling]
The product might explain:
Performance was below your recent baseline. This session followed shorter sleep, higher activity, and a longer period without food. These factors may have contributed, but one session does not prove causation.
It could then recommend:
Repeat the planned weight next session and test a carbohydrate-containing meal 60–120 minutes before training.
The goal is not to present the recommendation as a fact.
The goal is to turn the available evidence into a reasonable next test.
This project gives me a structured way to experiment with:
- Using AI to organize and explain personal health data
- Separating deterministic calculations from AI-generated language
- Designing prompts that use real user context
- Testing how useful AI-generated recommendations feel
- Understanding where AI is helpful and where it is unreliable
- Communicating uncertainty clearly
- Building responsible health-related AI experiences
The current project has several limitations:
- It begins with one primary user: me
- Some data is self-reported
- Strength workouts were recorded inconsistently in a note-based journal
- Exercise, set, repetition, load, and effort data are not complete enough for reliable longitudinal conclusions
- Historical health signals cannot be used to claim strength progression without a consistent performance outcome
- Personal patterns may not generalize to other people
- Observational relationships do not prove causation
- AI-generated explanations can be wrong
- Recommendations require further testing and validation
These limitations are part of the project, not something I want to hide.
This repository is intentionally built in public. Rather than generating synthetic long-term results, each iteration reflects real product development, real data collection, and continuous refinement of the measurement system.
Below is a current mobile interface prototype and UI/UX direction. Currently still in development and planning. This serves as a rough idea of how a future established app would look and feel for this intended purpose.
