Projects

Production systems, systems-programming work, and research engineering — each built with real constraints and real impact metrics.

All Projects

Distributed Task Queue System

40K+ daily jobs · P99 < 200ms3 repos

High-throughput async job processor with priority scheduling, retry with exponential backoff, dead-letter queues, and tenant isolation.

3 private reposCase Study
GoRedisKubernetes

Real-time Event Streaming Pipeline

5M+ events/day · <100ms latency3 repos

Kafka-backed audit logging and analytics pipeline with schema registry, consumer groups, and ClickHouse OLAP sink.

3 private reposCase Study
KafkaGoClickHouse

API Gateway & Rate Limiting Layer

30K+ req/day · multi-tenant3 repos

Token-bucket rate limiting backed by atomic Redis Lua scripts. Circuit breaking + distributed tracing via OpenTelemetry.

3 private reposCase Study
NginxRedisOpenTelemetry

LSM-Tree Key-Value Store

100K+ write ops/sec1 repo

From-scratch LSM-tree storage engine in Go with WAL, memtable, SSTable compaction, and Bloom filter.

GoStorage EnginesLSM-Tree

Consistent Hashing Load Balancer

< 5% key redistribution1 repo

Virtual-node hash ring with health checking, weight-based placement, and automatic rebalancing.

GoDistributed SystemsLoad Balancing

BanglaVQA Dataset Pipeline

20K+ annotated QA pairs2 repos

Automated collection, annotation tooling, and quality filtering for the Bangla VQA benchmark.

PythonHuggingFaceCLIP