14. Tech stack recommendation
Sub-pages
Section titled “Sub-pages”| Page | Covers |
|---|---|
| 14.1 Backend services | Spring Boot, modules, libraries, API gateway |
| 14.2 Data and storage | PostgreSQL, object storage, search, warehouse, ETL |
| 14.3 Workflow and rules | Workflow engine, rule engine, event bus, queue |
| 14.4 Infrastructure and deployment | Cloud, K8s, IaC, CI/CD, observability |
| 14.5 Security and compliance | IAM, secrets, encryption, audit, DR / BCP |
Stack at a glance
Section titled “Stack at a glance”| Layer | Choice |
|---|---|
| Language | Java 17 LTS (or 21) |
| Framework | Spring Boot 3.x |
| DB | PostgreSQL 15+ (managed: RDS / Aurora) |
| Object storage | S3 (or Azure Blob / OCI Object Storage) |
| Cache | Redis (ElastiCache) |
| Event bus | RabbitMQ (MVP) → Kafka / MSK (scale) |
| Workflow | Camunda 7 (MVP) → Temporal (scale) |
| Rule engine | JSON Logic + decision tables (MVP) → Drools (scale) |
| Search | OpenSearch |
| Warehouse | PostgreSQL replica (MVP) → Snowflake or ClickHouse |
| ETL | dbt; CDC via Debezium |
| BI | Metabase (MVP) → Superset / Looker (scale) |
| Cloud | AWS Mumbai primary, Hyderabad DR |
| Orchestration | EKS (Kubernetes) |
| IaC | Terraform |
| CI / CD | GitHub Actions / GitLab CI / Jenkins |
| Secrets | AWS Secrets Manager + KMS (or HashiCorp Vault) |
| Observability | Datadog or Grafana + Prometheus + Loki + Tempo |
| Identity (internal) | Google Workspace / Microsoft Entra ID + Okta |
| Identity (borrower) | Mobile OTP + Aadhaar |
| Encryption | TLS 1.2+ in transit; AES-256 at rest with KMS keys |
| ML platform | Feast + MLflow (at year 2+) |
| Frontend | React + Vite + TypeScript |
| Mobile | React Native or native (if mobile app needed) |
| Field-agent app | React Native (offline-first) |
Why this stack
Section titled “Why this stack”- Java / Spring Boot: team familiarity, fintech industry standard, deep library ecosystem, strong ops tooling.
- PostgreSQL: ACID for finance; rich extensions; logical replication for CDC.
- Camunda / Temporal: long-running workflows are first-class.
- JSON Logic + decision tables: rules editable by non-engineers; promoted to Drools when complexity demands.
- Kafka / RabbitMQ: event-driven async without distributed-tx pain.
- AWS: best India-region availability, broadest managed services.
See sub-pages for detail.