Anchita
A cloud-native platform for the institutional banking client lifecycle — onboarding, servicing, and payment processing, orchestrated end to end.
Architecture & design documentation — implementation in progress.
⚠️ Fictional reference architecture — This is a personal learning project. The platform, all business data, system names, and product counts are entirely fictional. This documentation does not describe any real employer, client, production environment, or proprietary implementation.
A personal portfolio project — built to demonstrate how modern cloud-native engineering, durable workflow orchestration, and agentic AI can replace legacy workflow complexity in institutional client onboarding and payments.
Platform Overview Architecture
What Is Anchita?
Anchita is a cloud-native platform for the institutional banking client lifecycle and payments designed from first principles for commercial banking. It serves bankers, operators, compliance officers, and institutional clients across all business segments — Small Scale, Medium Scale, and Enterprise — through a single coherent workflow engine, one unified data model, and one consistent operator experience.
Every operator on the Anchita platform has an AI co-pilot at every step of the servicing workflow.
Platform at a Glance
| Architecture | Six-layer cloud-native system on Amazon EKS |
| Workflow engine | Temporal.io — durable, long-running, signal-driven |
| Main workflow | 8 phases, segment-aware routing, child workflow hierarchy |
| Products | 20+ treasury products — each an independent Temporal child workflow |
| AI agents | 6 specialised agents + conversational AI Assistant panel |
| Integrations | Multiple external systems — one dedicated Spring Boot adapter each |
| Client segments | SMALL · MEDIUM · ENTERPRISE — one workflow, segment-aware routing |
The Problem It Solves
Commercial banks today onboard treasury clients through ageing BPMN workflow platforms that were designed for a different era:
- Fragmented operator experience — a different wizard screen for every step, no unified case view
- No intelligence — operators manually check documents, fill AML questionnaires, and look up client data from separate systems
- Brittle integrations — a single integration gateway handling all external system calls with no resilience patterns
- Segment silos — Small, Medium, and Enterprise clients handled by entirely separate process definitions
- No durability — long-running cases spanning days to weeks are prone to failure with no recovery story
- Compliance risk — manual steps with no structured audit trail for BSA/AML (Bank Secrecy Act / Anti-Money Laundering) review decisions
Anchita replaces all of this with durable workflow orchestration, event-driven integration, and agentic AI assistance — designed with the compliance requirements of a regulated banking environment as first-class constraints.
Architecture at a Glance
┌─────────────────────────────────────────────────────────────────┐
│ LAYER 1 — PRESENTATION │
│ React Web (Operator Workspace) · React Native (Banker Mobile)│
│ AI Assistant Panel (embedded, role-aware) │
├─────────────────────────────────────────────────────────────────┤
│ LAYER 2 — API GATEWAY │
│ Spring Boot — GraphQL BFF (workspace) │
│ Spring Boot — REST API (mobile + external channel intake) │
├─────────────────────────────────────────────────────────────────┤
│ LAYER 3 — WORKFLOW ORCHESTRATION │
│ Temporal.io — durable, long-running, signal-driven workflows │
│ ClientServicingWorkflow + child workflow hierarchy │
├─────────────────────────────────────────────────────────────────┤
│ LAYER 4 — AGENTIC AI │
│ Python / FastAPI agent service │
│ DocumentReviewAgent · CompliancePreFillAgent · AI Assistant │
├─────────────────────────────────────────────────────────────────┤
│ LAYER 5 — INTEGRATION SERVICES │
│ Spring Boot adapters — one per external system (multiple) │
│ Circuit breakers · Retry policies · OpenAPI 3.0 contracts │
├─────────────────────────────────────────────────────────────────┤
│ LAYER 6 — DATA │
│ Amazon Aurora PostgreSQL · Amazon S3 · Amazon ElastiCache │
└─────────────────────────────────────────────────────────────────┘
All layers on Amazon EKS
Browse the Documentation
| Section | What’s Inside |
|---|---|
| Platform Overview | Design goals, core architectural decisions, and the reasoning behind Anchita |
| Architecture | Technology stack decisions + deep-dive into each of the six architecture layers |
| Workflow Design | The Temporal workflow hierarchy — phases, child workflows, product fan-out, all Mermaid diagrams |
| Security & Planning | Security architecture, compliance boundaries, and open architectural decisions |
Technology Highlights
| Area | Choices |
|---|---|
| Frontend | React + TypeScript · React Native · Ant Design |
| Backend APIs | Spring Boot 3.x · GraphQL (Netflix DGS) · REST (OpenAPI 3.0) |
| Workflow engine | Temporal.io · Java 25 LTS · Virtual threads (Project Loom) |
| AI agents | Python 3.x · FastAPI · LLM provider TBD (AWS Bedrock preferred) |
| Integration | Spring Boot adapters · Resilience4j circuit breaker · OpenAPI 3.0 per system |
| Go | anchita-cli operations CLI · WebSocket notification hub |
| Data | Amazon Aurora PostgreSQL (pgvector) · Amazon S3 · Amazon ElastiCache (Redis) |
| Infrastructure | Amazon EKS · KEDA event-driven autoscaling · AWS App Mesh · ArgoCD GitOps |
Deliberate Polyglot Architecture
Each language is chosen for a specific reason — not for variety:
| Language | Where | Why |
|---|---|---|
| Java 25 | Temporal workflows · Spring Boot APIs · Integration adapters | Deep ecosystem (Resilience4j, DGS, OpenAPI generator) · Virtual threads for Temporal worker concurrency |
| Python 3 | AI agent service · LLM toolchain | Native ML/AI ecosystem · LangChain / LangGraph / Anthropic SDK |
| Go | anchita-cli · Notification Service |
Single static binary for CLI · Goroutine model for 10k+ concurrent WebSocket connections at minimal RAM |
Fictional reference architecture — personal learning project exploring cloud-native engineering patterns in a regulated banking domain.