Case Studies

Architectural deep dives into production-grade systems and AI infrastructure.

1. FamCARE: Hyper-Local Service Marketplace Architecture

An in-depth analysis of how we built a modular monolith using FastAPI and PostgreSQL to coordinate customers, caregivers, and managers.

• Ghost Assignment Prevention: Implemented atomic proposals and Redis-backed reservations with 180-second TTL cleanup task schedulers.
• Real-Time Broadcasts: Built a WebSocket system backed by a Redis Pub/Sub backplane to synchronize state changes across clients in under 300ms.
• Stress Testing: Benchmarked the system to handle 2,000+ requests/minute under 100 concurrent operational slots.

2. shoppin': Large Scale Multimodal Search Infrastructure

How we scaled ML ingestion and retrieval to serve 30-40 million fashion catalog items.

• Multimodal CLIP Retrieval: Deployed CLIP/VLLM vector index pipelines on AWS SageMaker with Redis cache overlays.
• GitOps & Rancher: Configured AWS EKS clusters using Terraform, enabling automated failover and zero-downtime deployment pipelines.

3. WHAM! OTT: Hybrid Video Stream Engine

Engineering a hybrid video playback client with comic-book UI layouts and postMessage sync layers.