โจ Feature Demonstrations
12-Week Program Engine
Phase Rules
- Weeks 1โ2: 2 sets at 3โ4 reps in reserve (RIR)
- Weeks 3โ4: the listed sets at 2โ3 RIR
- Weeks 5โ12: 1โ2 RIR, with a final isolation set near failure allowed in weeks 9โ12
- Double progression suggests the next weight or rep target from what was logged last time
- Warm-up ramps go to the day's first compound lift, unless the plan sets them per exercise
Weekly Mix & Exercise Randomizer
How the Randomizer Chooses
- The replacement trains the same main muscle and is the same kind of exercise: compound for compound, isolation for isolation, hold for hold
- It must move in much the same way, and never brings in a supporting muscle from the other half of the body
- Candidates matching the original's one-sidedness, or its bodyweight vs. loaded style, are twice as likely
- Each extra supporting muscle a candidate adds halves its chance
Offline-First Workout Logging
Outbox Flow
- Every queued write is idempotent: a PUT by an ID made on the device, or a DELETE, so retrying is always safe
- A newer write with the same key replaces or merges into the queued one, so only the latest state is sent
- Writes are sent in order, stopping at the first one that can't reach the server yet
- Flushing runs on reconnect, on returning to the foreground, and every 30 seconds
- Screens show which entries are "saved on this phone" until they reach the server
PUT /api/sessions/{id}. Even a caption with blocked language is dropped rather than rejected, so a workout saved offline can never fail to sync because of its caption.
Nutrition, Body & Health Tracking
Targets & Imports
- Maintenance calories from Mifflin-St Jeor × activity, minus a small deficit
- Protein about 1 g per lb, fat about 0.36 g per lb, carbs fill the rest
- When the app opens, the last 14 days of steps, sleep and weigh-ins are imported from the phone's health app
- Values typed in by hand always win over imported ones, and weigh-ins are de-duplicated by sample ID
- Progress photos are stored in Cloudflare R2 and can be compared side by side
Scored Weekly Check-Ins
Habit Items
- Hit the protein target on at least 6 days
- Stayed near the calorie target: the average of logged days within tolerance, with most days logged
- Met the step and cardio goal: average steps at or above the goal, plus two cardio sessions of 20+ minutes
- Averaged at least 7 hours of sleep over the logged nights
- "Following the plan" means about three-quarters of the checklist (6 of 8 for the upper/lower plan)
Friends-Only Social, Leaderboards & Moderation
Fair Rankings
- Leaderboards rank only things that are fair across body sizes: checklist adherence, workouts, PRs, and strength gained relative to your own start
- Adherence is ranked by the share completed, since checklists differ in length by workout style
- Strength gain is the average percent change in each exercise's best estimated one-rep max (Epley) between a baseline window and a recent one
- Absolute volume is deliberately not ranked
- Members can report and block; the owner is alerted right away so reports are handled within a day, as App Store guideline 1.2 requires
๐ Professional Summary
The backend is an ASP.NET Core API on .NET 10 with EF Core and PostgreSQL. It holds the domain logic: phase rules, double-progression prescriptions, a seeded weekly mix, a constraint-based exercise randomizer, check-in scoring, nutrition target calculation and fair leaderboards. Integrations include Cloudflare R2 for photos, Resend for email and Expo push notifications, and the API is deployed with Docker on Railway.
The mobile app uses Expo SDK 57, React Native 0.86 and React 19 with Expo Router and TanStack Query, with a persisted outbox for offline-first logging and Apple Health / Health Connect imports. More than 120 xUnit tests cover rules, auth, training, tracking, social features and a security audit, run against a real PostgreSQL database.
๐ฏ Most Marketable Skills
Backend & Domain Logic
- ASP.NET Core on .NET 10
- EF Core 10 with PostgreSQL
- Rules engines for programs and check-ins
- Seeded, stable randomization
- JWT auth with BCrypt hashing
Mobile & Offline
- Expo SDK 57 / React Native 0.86
- Offline-first outbox with idempotent sync
- TanStack Query with a persisted cache
- HealthKit and Health Connect imports
- iOS, Android and web from one codebase
Product & Social
- Friends-only feed and plan sharing
- Leaderboards fair across body sizes
- Reporting, blocking and owner tools
- Push reminders and email
- Privacy by design (body data never shared)
Quality & Operations
- 120+ xUnit tests on real PostgreSQL
- Security audit test suite
- Docker deployment on Railway
- Cloudflare R2 object storage
- EAS builds for the app stores
๐๏ธ Architecture
Mobile & Web App
- Expo Router (file-based)
- Today, Train, Food, Progress, Friends tabs
- Onboarding and plan builder
- Photo compare and charts
- Owner admin screens
Data & Sync
- TanStack Query (persisted)
- Offline outbox
- NetInfo reconnect flushing
- Health app import
- Secure token storage
API
- Account and auth
- Training (sessions, plans)
- Tracking (food, body, activity)
- Social (feed, friends, boards)
- Admin and moderation
Domain Rules
- ProgramRules and BlockPlan
- PrescriptionService
- WeeklyMix and ExerciseRandomizer
- CheckInRules and TargetCalculator
- PrRules (Epley 1RM)
Infrastructure
- PostgreSQL via EF Core
- Cloudflare R2 (photos)
- Resend (email)
- Expo Push
- Docker on Railway
๐ Key Achievements
Never Loses a Set
An idempotent, ordered outbox keeps workouts safe through dead gym connections and syncs them automatically when the phone is back online.
Program, Not Just a Log
Phase rules, double progression, warm-ups and a seeded weekly mix turn a 12-week plan into a ready-to-go workout every day.
Smart Exercise Swaps
A constraint-based randomizer swaps in exercises that train the same muscle the same way, weighted toward the closest matches.
Fair Leaderboards
Friends compete on adherence, PRs and strength gained relative to their own start, so bigger lifters don't win by default.
Health Integrations
Steps, sleep and weigh-ins import from Apple Health and Health Connect, without ever overwriting values entered by hand.
Tested on Real PostgreSQL
More than 120 integration tests, including a security audit, run against a real PostgreSQL database that is recreated on each run.