DesiHub Tech Stack 2026: Building a Modern Banking Platform

Published: January 18, 2026

---

We're excited to share DesiHub's complete technology stack for 2026 as we scale our banking application to serve 50,000+ daily users. This article outlines our architectural decisions, tooling choices, and the reasoning behind them.

Executive Summary

DesiHub is building a modern banking platform with microservices architecture, focusing on security, scalability, and cost-effectiveness. Our stack leverages battle-tested open-source technologies combined with strategic paid services.

Target Metrics 2026:

• 50,000+ daily active users
• 10M+ API requests per day
• 99.99% uptime SLA
• Sub-100ms API response times

Our Technology Decisions

Backend Architecture: Go + Microservices

We chose Golang (v1.24+) as our primary backend language.

Why Go?

• Performance: Near C++ speeds with garbage collection.
• Concurrency: Native goroutines handle thousands of concurrent requests efficiently.
• Banking industry adoption: Used by Monzo, Nubank, and American Express.

Core Backend Stack:

Framework:      Gin (HTTP routing) & Fiber (for ultra-high perf services)
ORM:            GORM & sqlc (for type-safe SQL)
Auth:           golang-jwt/jwt + Keycloak
Authorization:  Open Policy Agent (OPA)

Frontend: Next.js 15+ with App Router

Frontend Stack:

Framework:        Next.js 15 (React Server Components, Partial Prerendering)
State Management: TanStack Query v5
Styling:          Tailwind CSS v4 + Shadcn/ui
Forms:            React Hook Form + Zod
Authentication:   Auth.js (formerly NextAuth)

Why Next.js 15?

• Partial Prerendering (PPR): Instant static shell with dynamic holes. Best of both worlds.
• Server Actions: simplified data mutations without separate API routes for simple tasks.

Database Architecture

Primary Database: PostgreSQL 17 We upgraded to Postgres 17 for its improved vacuuming and JSON performance.

Configuration Strategy:

• Master-replica setup for read scaling
• PgBouncer for connection pooling
• Neon (Serverless Postgres) for branching and dev environments

Caching Layer: Redis & DragonflyDB We are experimenting with DragonflyDB as a drop-in Redis replacement for higher throughput.

Message Queue: RabbitMQ & NATS JetStream

For asynchronous processing:

• RabbitMQ: For traditional reliable task queues (email, PDF generation).
• NATS JetStream: For high-throughput event streaming (transaction logs).

Infrastructure & Hosting

Hosting Provider: Hetzner Cloud & AWS Hybrid

• Hetzner: For compute-heavy workloads (saving 60% vs AWS).
• AWS: For managed services like S3 (blob storage) and RDS (when managed DB is needed).

Cost Breakdown (MVP Production):

Production Environment:
├── 3x App Servers (Hetzner AX52)  - €150/month
├── Managed Postgres (Aiven/Neon)  - $200/month
├── Load Balancer & WAF            - $50/month
Total: ~$400/month

DevOps & CI/CD

CI/CD: GitHub Actions Our pipeline now includes AI Code Review agents that pre-scan PRs for security vulnerabilities before a human reviews them.

Containerization:

• Docker for local dev.
• Kubernetes (k3s) for production orchestration.

Developer Desktop Specifications 2026

To maintain high velocity, every developer gets top-tier gear.

Standard Issue Workstation:

• CPU: Apple M4 Pro / AMD Ryzen 9 9950X
• RAM: 64GB DDR5 (Minimum)
• GPU: RTX 5070 (for local LLM inference)
• Monitors: Dual 4K 32"

Why 64GB RAM? Developers run the entire microservices stack + local LLMs (Llama 3 8B) for coding assistance locally.

Security Infrastructure

Authentication:

• Passwordless login (Passkeys) as primary method.
• Biometric step-up for high-value transactions.

Compliance:

• PCI DSS Level 1
• GDPR & PDPA (Uganda)

Future Roadmap

Q1 2026:

• Complete migration to Next.js 15.
• Roll out "Voice Banking" features using OpenAI Realtime API.

Q2 2026:

• Blockchain integration for Cross-border Stablecoin settlements.

---

Summary

Our 2026 stack is designed for speed, developer happiness, and financial reliability. By betting on Go, Postgres, and Next.js, we have a foundation that can scale to millions of users.