TimeUp

Operational SaaS for the health and wellness sector.

Summary

The Challenge

Local businesses operating blindly with WhatsApp and notebooks. Owners permanently busy and an operative staff with low affinity for complex technological tools.

The Solution

A Multi-tenant Cloud platform with friendly onboarding, biometric Passkeys without cumbersome passwords and a kinetic Real-Time flow.

The Impact

MVP built in 10 weeks under an optimized architecture. Deployment with near-zero operating costs before gaining real production traction.

RoleSolo Developer / Architect
Timeline10 Weeks (MVP)
ArchitectureServerless + Stateful VPS
Core StackNext.js, Supabase, Passkeys
Public AppBusiness Panel

"If onboarding took more than ten minutes, the implementation failed."

The Functional Premise

01 // Inflexible Constraints

Retaining walls.

The context that defined the architecture and forced us to avoid over-engineering.

Low Patience Users

Business owners aged 30 to 50. The flow could not assume any prior technical knowledge. That's why we transitioned from slow email schemes to using Native Biometric Hardware (Passkeys).

MVP Against the Clock

Developed and modeled by a single maintainer, forcing the choice of hyper-productive tools (Turborepo + Next).

Tenant Quality

Semi-open onboarding system. Requires administrative approval in the Backend, creating complex database states PENDING.

Serverless Zero-Cost

The app had to pivot with a dead burn rate. Pure deployment on Vercel & Supabase Free isolating demanding components.

Strict Legal Module

Non-negotiable management of Digital Informed Consents with PDF electronic signatures for invasive medical treatments.

02 // Ecosystem Interfaces

Dedicated tools by role.

Administrador TimeUp

MODULE : ADMIN

Global Analytics and Control Center.

Expand
Finanzas DueƱo

MODULE : OWNER

Performance Metrics and Configuration.

Expand
Agenda Staff

MODULE : STAFF

Live Agenda and Client-Local Management.

Expand

03 // Architecture & Decisions

Trade-offs that give life to the business.

3.1 / REPOSITORY STRUCTURE

Turborepo Monorepo

Building multiple React ecosystems forced crossing business logic, TypeScript types, and Prisma schemas. We opted for the initial rigidity of monorepositories to orchestrate multiple coupled bases with automatic caching.

The Operational Cost

Initial tooling delayed v.1, but scaling between deployments became completely instantaneous weeks later.

3.2 / STATEFUL SYNCHRONIZATION

WebSocket Isolation

Platforms like Vercel and Edge Functions punish prolonged open connections (Timeouts and dead SSE in minutes). We extracted the bidirectional engine to a Virtual Machine with NodeJS + pm2 + Redis PubSub acting as a real-time pipeline.

The Operational Cost

Monitoring fragmentation; we went from just looking at Vercel logs to having to constantly review metrics on an Ubuntu droplet.

3.3 / AUTHENTICATION EXPERIENCE

Biometric WebAuthn Login

Forcing local staff to remember long strings interrupts their operational work. The strong implementation of hardware-integrated biometric APIs (FaceID / Fingerprint) cut access friction from several minutes to fractions of a second on the counter tablet.

The Operational Cost

Extremely high complexity in the fallback; forced maintaining an extra robust system via SMTP Magic Links.

04 // Post-Mortem System

Raw lessons upon hitting Production.

Pure technical implementations rarely survive the organic behavior of the end user and lower-tier Cloud limitations. What actually caught fire and how it was shielded.

Connection Pooler Choking

Combining Serverless AWS with cold Database engines in Supabase ended up burning inefficient connection limits, destroying the pipelines.

URGENTLY MIGRATED TO ACTIVE TRANSACTION POOLERS

WebAuthn & Origin Mismatch

An insignificant extra slash '/' in the production CORS config collapsed the entire digital security key flow on the initial deploy because they break cryptographic security.

MANDATORY MIRRORING OF CI/CD PREVIEW ENVIRONMENT

State Purgatories & Ghost Users

Semi-approved entities created inconsistencies by dragging values that contaminated metrics of subsequent cohorts in KPIs.

ABSOLUTE RESTRICTION OF INTERMEDIATE STATES (BOOLEANS TO THE EXTREME)
LAST UPDATED: Q1 2026 // TIMEUP SYSTEM ARCHITECTURE LOGSTATUS: IN CONSTANT PRODUCTION