Our Tech Stack & Why We Chose It

Our Tech Stack & Why We Chose It

We optimize for developer speed, operational reliability, and security by design. Our default stack:

• Backend: NestJS (TypeScript)
• Database: PostgreSQL (managed on Azure)
• Caching & Queues: Redis (Azure Cache for Redis)
• Web: React (Vite or Next.js when SSR/SEO is needed)
• Mobile: React Native (Expo first)
• Cloud: Microsoft Azure (Container Apps/AKS, Key Vault, App Gateway/Front Door, Application Insights)

Architecture at a glance

• Clean, modular services in NestJS (monolith-first or modular monolith → services when needed).
• PostgreSQL as the system of record; Redis for read-through cache, rate limits, ephemeral sessions, and BullMQ queues.
• HTTP APIs (REST/OpenAPI) and background workers; webhooks for integrations.
• React web app + React Native (Expo) mobile from a monorepo (shared UI/types).
• Azure: Containerized workloads, Key Vault for secrets, Application Insights + OpenTelemetry.

Backend — NestJS

Why NestJS? Opinionated structure, DI, testability, and first-class TypeScript.

Database — PostgreSQL

Why Postgres? Reliability, strong SQL + JSONB, great extensions (pgcrypto, trigram).

Azure choice

• Azure Database for PostgreSQL – Flexible Server, zone-redundant HA where required.

Data access

• Sequelize with explicit migrations and models
• Use of idempotent migrations, forward-only; verify in staging.

Caching & Queues — Redis

Use cases

• Read-through cache for hot queries (TTL 5–30 min).
• Idempotency keys, rate limits, ephemeral sessions.
• BullMQ queues for background jobs (emails, webhooks, ETL), with backoff/retry.

Web — React

Why React? Ecosystem depth; smooth handoff to React Native.

• Vite for SPAs; Next.js when SSR/SEO or edge rendering helps.
• UI: headless components or design system.
• State: Zustand with tanstack React Query; forms via React Hook Form + Zod.

Mobile — React Native (Expo)

Why Expo? Rapid iteration and great device APIs.

• Offline-first with SQLite where necessary
• Shared logic with web via monorepo libs (types, validation, API SDK).

Cloud — Azure

Compute

• Azure Container Apps (simple) or AKS (at scale).
• Background workers as separate containers; autoscale on queue length or CPU.

Networking & delivery

• Azure Front Door or Application Gateway for WAF + TLS; CDN for assets.

Secrets & identity

• Azure Key Vault (managed identities for access).
• Azure AD / Entra ID or B2C for auth; JWTs to apps; RBAC in services.

Data & storage

• Azure PostgreSQL Flexible Server, Azure Cache for Redis, Blob Storage for uploads; lifecycle rules for retention.

Observability

• Application Insights with Prometheus export (traces, logs, metrics).
• Logging with Loki; streamed to Grafana

Tooling

• ESLint/Prettier shared config
• Zod schemas shared across API and clients (type-safe DTOs).

Golden Defaults

PostgreSQL

• uuid primary keys; timestamps with time zone.
• Sequelize migrations; never hot-edit prod.
• Indices for foreign keys, unique business keys; partial indexes for common filters.

Redis

• TTL on all cache entries; never store PII.
• Separate DBs or prefixes for cache vs. queues.
• Dead-letter queues for jobs; alert on retries > N.

NestJS

• Global validation pipe (Zod or class-validator), strict DTOs.
• Central error filter → problem+json; consistent error codes.
• Feature flags (configurable via Azure App Configuration or env).

React / RN

• Error boundaries; analytics with consent; i18n scaffolding.
• API SDK generated from OpenAPI (type-safe).
• Form validation with shared Zod schemas.