Global, high-traffic media platform serving multiple brands with personalization, CMS, BFF services, and compliance-critical workflows.
- Built Node.js/TypeScript microservices for an in-house content platform powering Food Network, CNN, Eurosport, Magnolia, Golf Digest, and MotorTrend.
- Architected AWS-native infrastructure (Lambda, ECS, SQS, SNS) with Infrastructure-as-Code; designed Lambda-vs-ECS deployment trade-offs per service.
- Delivered a GraphQL Backend-for-Frontend for Food Network with auth, observability, and proactive monitoring.
- Integrated OneTrust to enable GDPR/CCPA user-data access, deletion, and auditability across 6+ services.
- Standardized REST and GraphQL API conventions adopted across teams; cut cross-team onboarding friction.
- Introduced distributed tracing (OpenTelemetry) and Redis caching, reducing DB pressure and improving incident MTTR.
Two of the systems I built here are walked through with architecture diagrams in the section below.
High-volume, event-driven integrations platform connecting POS systems, marketing, and user segmentation pipelines.
- Designed and built an event-driven integration platform connecting 10+ third-party POS systems using Kafka and Redis.
- Reduced transaction latency by ~50% through async pipelines, cache-aside, and partition-aware consumer scaling.
- Built customer segmentation on MongoDB + ElasticSearch, driving ~25% uplift in user conversions and ~15% bounce-rate reduction.
- Delivered three enterprise-grade microservices contributing to ~25% YoY revenue growth.
- Owned Kafka topic schemas and backward-compatible event evolution strategy.
Fintech P2P lending platform with automation, risk assessment, and compliance requirements.
- Built a P2P lending platform on Express.js + MongoDB + AWS; automated ~95% of credit workflows.
- Improved loan approval accuracy by ~25% via system-driven decision pipelines; cut end-to-end processing time by ~50%.
- Implemented OCR + facial recognition for KYC and an NLP-based scoring engine for risk analysis.
- Reduced communication errors by ~90%, lifting customer satisfaction by ~20%.
Modular, multi-tenant SaaS budgeting platform. Ingests bank statements, enriches transactions via rule-based + AI categorisation, and serves a tenant-aware GraphQL API. Built as a pnpm monorepo on AWS serverless.
- 3 Lambda services + 8 DynamoDB tables fully provisioned via AWS CDK; on-demand pricing throughout.
- Fully event-driven ingestion: S3 → SQS → txn-loaders Lambda → DynamoDB Stream → tag-loaders Lambda — zero polling.
- Multi-tenant isolation via
tenantId partition key across all tables; JWT auth via SSM Parameter Store. - Hybrid categorisation: deterministic NLP rule engine first; Amazon Bedrock (Mistral) fallback for low-confidence matches — cost-optimised.
- GraphQL API with Apollo Server 5 — 7 queries, 18+ mutations; auth (incl. refresh-token rotation), two-step uploads, financial reviews, forecasts, budgets, savings goals, sinking funds.
- Observability: AWS X-Ray (10% sampling), CloudWatch alarms (errors, duration, throttle), Winston structured logging, Prometheus
/metrics endpoint.
Architecture diagram below ↓