Backend Software Engineer · Distributed Systems · Database Internals · Performance

Building to learn, evolve, and scale: I work on backend systems so their behavior under real conditions stays clear—architecture, failure modes, and performance. Focused on distributed systems, database internals, and high-performance backends, with clear design and explicit trade-offs.

I write about what I learn on Hashnode—lessons from production work, side projects, databases, and the systems behind them.

• Distributed systems: event-driven design, Kafka, messaging, consistency, and fault tolerance
• Data systems: relational modeling, query execution, indexing, OLAP, analytical pipelines, and Neo4j/graph data
• Backend performance: caching, queues, concurrency, benchmarking, and load testing
• Operability: observability, failure drills, containers, and delivery automation

Stack: TypeScript · Node.js · NestJS · PostgreSQL · Redis · Kafka · DuckDB · Neo4j · Docker · Prometheus/Grafana

Selected projects — newest first.

A real-time ride-sharing GPS pipeline covering Kafka ingestion, stream processing, observability, and live client delivery.

Technical highlight: Built a three-broker KRaft cluster with Avro schema evolution, transactional ETA processing, ksqlDB anomaly detection, Prometheus/Grafana monitoring, and Redis GEO/Pub/Sub fan-out through Socket.IO.

A Formula 1 OLAP engine that ingests real race data into a Parquet lake and serves analytics through NestJS and DuckDB.

Technical highlight: Modeled independent-grain facts with a Kimball-style star schema and built benchmarks that compare partition pruning and raw aggregation against materialized summaries.

A Graph RAG knowledge base that turns research papers into connected, queryable evidence for AI assistants.

Technical highlight: Built an end-to-end ArXiv ingestion pipeline for sections, citations, concepts, and embeddings, then exposed Neo4j graph and vector retrieval through five MCP tools over stdio and HTTP.

A persistent, zero-infrastructure task queue for Node.js, distributed as @km-dev/lite-q.

Technical highlight: Implemented a SQLite WAL-backed job state machine with atomic claims, delayed and cron scheduling, exponential retries, Prometheus metrics, and separate concurrency paths for I/O handlers and worker-thread jobs.

A production-style NestJS API with advanced filtering, cursor pagination, Redis caching, containerized startup, and CI.

Technical highlight: Tested 100,000 seeded records with 50 concurrent k6 users, achieving 55.4 requests/second, 6.58 ms p95 latency, and a 0% error rate on a local development machine.

Also: Competitive Programming — Java and C++ solutions from Codeforces.

I keep structured notes on distributed systems, databases, AWS, backend engineering, and interview problems in my-study-notes.

On Hashnode I write about how I learn—lessons from production-grade work through my career, and from side projects, databases, networking, and the systems behind them.

Selected posts:

• When One Database Is No Longer Enough: A Practical Guide to Distributed Transactions
• Latency, Throughput & the Laws That Govern Every System Under Load
• What I Learned Building Real Observability Into an App

Open to collaborating on distributed systems, database internals, backend infrastructure, and developer tooling.