Case Study

End-to-end case studies on distributed systems and production architecture. Each one documents the problem, constraints, decisions, tradeoffs, and lessons learned.

Distributed System · Full Case Study

KaryaFlow — Distributed Job Queue

Problem: Multi-tenant async task processing at scale with exactly-once semantics.
Design: Token-bucket rate limiting, Redis Lua atomics, K8s horizontal scaling, dead-letter queues.
GoRedisKubernetes

Other Case Studies

Bangla VQA Research Pipeline

Problem: Scalable annotation and quality-control pipeline for multimodal Bangla datasets.
Design: Distributed crawlers, quality filters, human-in-the-loop annotation, CLIP embeddings.
PythonHuggingFaceCLIP

Event Streaming & Audit Log

Problem: Zero-loss audit pipeline for compliance-critical multi-service events.
Design: Kafka consumer groups, schema registry, exactly-once producers, ClickHouse OLAP.
KafkaGoClickHouse

API Gateway Rate Limiter

Problem: Per-tenant rate limiting across 30K+ req/day with sub-5ms overhead.
Design: Atomic Redis Lua scripts, token-bucket algorithm, circuit breaker, OpenTelemetry tracing.
RedisNginxOpenTelemetry