Tendon is a live fitness marketplace connecting users with personal trainers. I designed and built the entire backend from the ground up — from the first line of code to production deployment — and continue to maintain and expand it today.

What I Built

• Core API & Auth — Phone OTP, email/password, and social login (Google, Apple, Facebook) via Laravel Sanctum. User and trainer models.
• Nearby Trainer Discovery — Geo-aware search powered by Meilisearch with a composable query builder. Users filter by distance (geo radius), category, rating, price, gender, and time-of-day availability — all composed into a single search call per request. Results sort by proximity or trainer score. Pagination uses a buffer multiplier to account for post-search availability filtering, ensuring a full page every time. Moving this off the database entirely keeps results under 100ms even with multiple simultaneous filters.
• Trainer Profiles — Photo galleries, skills, and verified identity documents (Aadhaar, PAN, UPI) via Surepass and Digilocker APIs. Admin approval workflow before trainers go live.
• Booking System — Real-time timeslot availability, solo/family/co-booking, and installment-based payment tracking.
• Client Progress & Workout Tracking — When a client books a trainer, the system automatically generates a personalized multi-step journey based on the trainer’s configured workflow. Steps cover health profile collection, lifestyle assessment, fitness evaluation, goal setting, workout planning, and live workout tracking. Each step type has its own executor via a registry pattern — adding new step types doesn’t touch existing code. The workout step supports viewing a session calendar, starting/completing sessions, marking individual sets and reps in real time, rescheduling, and cancelling — all within the same step. Sessions capture an RPE score and notes on completion. Form-based steps (assessments) are linked to the questionnaire system; on submission, an event-driven listener chain auto-advances the client to the next step with no manual orchestration. Every step transition is idempotent — re-triggering the same event never creates duplicate records or double-fires side effects. Clients sharing an account (e.g. family members) each get fully independent progress records.
• Custom Questionnaires — Trainers create, clone, and assign questionnaire templates; responses feed into progress workflows automatically.
• Ratings & Reviews — Nested replies, likes, and automatic trainer tier recalculation based on ratings and approval status.
• Admin Panel — Built with Filament v3; trainer approvals, document verification, user reports, and role-based access control.
• CI/CD Pipeline — Set up end-to-end deployment automation on AWS using CodePipeline and CloudFormation. Changes pushed to the main branch automatically build, push to ECR, and deploy to ECS — infrastructure defined entirely as code.
• Notifications & Integrations — AWS SNS (SMS), AWS SES (email), AWS S3 (storage), Sentry (error tracking).

By the Numbers

• 286 database migrations across a schema of 67 models
• Versioned API (V1 + V2) with 49 controllers serving mobile and web clients
• 57 test files covering features, integrations, and API endpoints

Stack: PHP/Laravel · PostgreSQL · Redis · AWS · Meilisearch