Back to Portfolio

Case Studies

Deep-dive into production systems I've designed and built — architecture decisions, technical challenges, and measurable outcomes.

GraphQL Federated Supergraph

Transnational Software Services

Designed and built a Netflix-style GraphQL Federated Gateway that unifies 10+ domain microservices into a single, queryable supergraph — the backbone of a large-scale professional networking platform.

THE CHALLENGE

The platform had 10+ independently developed microservices, each with its own REST API. Frontend teams were making dozens of API calls per page, leading to slow response times, inconsistent data formats, and painful coordination between teams.

THE SOLUTION

  • Implemented Apollo Federation 2.0 with a centralized gateway that composes sub-graphs from each microservice
  • Used @key directives for entity resolution across services, enabling seamless data joining at query time
  • Built schema registry with automated validation to prevent breaking changes during deployments
  • Implemented DataLoader pattern to batch and deduplicate database queries, reducing N+1 problems
  • Added query depth limiting and complexity analysis to prevent abuse and ensure consistent performance
  • Set up distributed tracing with OpenTelemetry for end-to-end query performance visibility

RESULTS & IMPACT

API Calls

15-20 per page

1-2 per page

Response Time

800ms avg

120ms avg

Dev Velocity

Cross-team sync required

Independent sub-graph deploys

Schema Coverage

Fragmented REST docs

100% typed, auto-documented

ARCHITECTURE

→ Client → Apollo Gateway → Sub-graph Router
→ Sub-graphs: Core, Auth, Jobseeker, Employer, Payment, Internship, Apprenticeship
→ Each sub-graph: FastAPI + Strawberry GraphQL + PostgreSQL
→ Event Bus: SQS/SNS for async cross-service communication
→ Monitoring: Datadog + CloudWatch + OpenSearch

TECH STACK

PythonFastAPIApollo FederationGraphQLSchema RegistryDockerAWS ECSDataLoaderOpenTelemetry

Professional Networking Platform

Transnational Software Services

Sole backend engineer building an entire large-scale professional networking platform from scratch — real-time social feeds, messaging, authentication, and all domain services.

THE CHALLENGE

Building a production-grade professional networking platform with real-time features, secure authentication, bot prevention, and scalable session management — all as a single backend engineer owning the entire architecture.

THE SOLUTION

  • Designed microservice architecture with clear domain boundaries and event-driven communication via AWS EventBridge
  • Built WebSocket gateway for real-time features — live feed updates, instant messaging, typing indicators, and push notifications
  • Implemented reCAPTCHA v3 with action-based scoring for bot prevention, validating both frontend tokens and backend verification
  • Designed Redis-backed session management with dual persistence strategy — localStorage for 'Keep Me Signed In' and sessionStorage for standard sessions
  • Built comprehensive RBAC authorization layer with JWT + AWS Cognito integration
  • Implemented rate limiting with SlowAPI to protect endpoints from abuse while allowing legitimate high-frequency operations

RESULTS & IMPACT

Architecture

Monolithic concept

7+ domain microservices

Real-time

Polling-based

WebSocket with <50ms latency

Bot Prevention

None

reCAPTCHA v3 + rate limiting

Auth

Basic

Multi-layer JWT + Cognito + RBAC

ARCHITECTURE

→ API Gateway → FastAPI Microservices (7 domains)
→ WebSocket Server → Redis Pub/Sub → Connected Clients
→ Auth Flow: Cognito → JWT → RBAC Middleware → Resource
→ Session: Redis with configurable TTL (24h / 30d)
→ Events: EventBridge → SQS → Consumer Services

TECH STACK

PythonFastAPIWebSocketsRedisAWS CognitoJWTreCAPTCHA v3SlowAPIEventBridgePostgreSQL

ZYPE — Fintech Loan Platform

Respo Financial Capital

Developed backend microservices for a full fintech loan lifecycle — KYC verification, credit underwriting, and cash disbursement via banking APIs with real-time Kafka data streaming.

THE CHALLENGE

Building a secure, compliant fintech platform handling sensitive financial operations — KYC verification, credit scoring, and real-time cash disbursement through bank APIs — with strict data integrity and audit requirements.

THE SOLUTION

  • Built end-to-end loan lifecycle backend — application intake, KYC verification, credit underwriting, approval workflow, and disbursement
  • Integrated Yes Bank APIs for real-time cash disbursement with retry logic, idempotency keys, and transaction reconciliation
  • Implemented real-time data streaming from third-party LMS using Kafka on Confluent Cloud for loan status synchronization
  • Designed Kong API Gateway configuration for rate limiting, authentication, and request/response transformation
  • Built comprehensive API documentation with Swagger/OpenAPI for third-party integration partners
  • Implemented audit logging for all financial transactions with immutable event sourcing pattern

RESULTS & IMPACT

Loan Processing

Manual, days

Automated, minutes

Data Sync

Batch (daily)

Real-time via Kafka

API Docs

Informal

100% Swagger coverage

Disbursement

Manual transfer

Automated via Yes Bank API

ARCHITECTURE

→ Client → Kong API Gateway → Microservices
→ Loan Flow: Application → KYC → Credit Score → Approval → Disbursement
→ Kafka: LMS Events → Confluent Cloud → Consumer Services
→ Bank Integration: Yes Bank API with idempotency + retry
→ Audit: Event sourcing for all financial transactions

TECH STACK

Node.jsExpressKafkaConfluent CloudMySQLKong GatewaySwaggerYes Bank APIRedis

Interested in working together?

Get in Touch