Latest · October 2026
I Built a Distributed Task Queue from Scratch with Go and PostgreSQL
Solving the dual-write problem with transactional enqueueing (EnqueueTx), FOR UPDATE SKIP LOCKED, and debugging an 8x worker polling bottleneck with k6.
Building backend infrastructure, distributed systems, and database tooling from the ground up, working primarily with Go and PostgreSQL.
Writing
Projects
- Dhara: Distributed Task Queue in Go Go · PostgreSQL · pgx · Distributed Systems · k6
Distributed background job library and service backed by PostgreSQL. Solves the dual-write problem via transactional enqueueing (`EnqueueTx`), uses `FOR UPDATE SKIP LOCKED` for atomic worker claiming, and handles crash recovery via heartbeats and a background reaper (~148 tasks/sec under k6 load).
- Vow: PostgreSQL Migration Runner for Go Go · PostgreSQL · pgx · Database Tooling
Standalone SQL migration runner and CLI for `pgx/v5`, extracted from Dhara. Enforces schema immutability via SHA-256 checksum drift detection, uses PostgreSQL advisory locks for concurrent deploy safety, and embeds via `io/fs`.
- Volter: Distributed & In-Memory Rate Limiter Python · Redis · Lua · FastAPI · Concurrency PyPI
Published Python rate-limiting package (`pip install volter`) implementing Token Bucket and amortized O(1) Sliding Window Log algorithms. Uses two-level lock-free read paths in memory and atomic server-side Lua scripts in Redis, with FastAPI ASGI middleware.
- Redis Java: High-Performance Redis Server Java · Virtual Threads · TCP · RESP Protocol · async-profiler
Zero-dependency Redis server in Java using virtual threads. Profiled with `async-profiler` flame graphs to eliminate syscall and allocation bottlenecks, driving a 70x speedup to 84,000 ops/s across 36 commands, RESP protocol, pub/sub, and master-replica replication.