Claude Time: How Many Hours Do I Actually Spend in Claude Code?
An open-source macOS menu bar app and CLI that reads Claude Code's own transcripts to show time per project, skipping idle gaps and counting parallel sessions once.
Hello, world!
I'm Ozgur GurcanIzmir, TR
- I'm a Software Engineer focusing on Backend & Distributed Systems.
- I build high-throughput, resilient microservices.
- Primary tools: Java, Spring Boot, Go, and Kafka.
- Exploring AI agents, MCP tooling, and TypeScript.
- Based in Izmir, Turkiye.
- A builder, continuous learner, and freedom seeker.
- I love rock music and playing guitar.
- I love playing chess .
- I love playing video games , Dota 2 is my favorite.
- I'm Fulden's husband.
I specialize in backend engineering, high-throughput microservices, and distributed systems. Having worked on production systems handling millions of requests per minute, I created this space to document practical architectural lessons, zero-downtime database migrations, and performance tuning.
Beyond backend infrastructure, I enjoy building side projects, experimenting with AI workflows (MCP servers, agents, and modern tooling), and self-hosting. This blog is my digital workbench for sharing engineering notes that solve hard, real-world problems.
Technical Articles & Case StudiesSide Projects & ToolingResume & Career JourneyLive Website Analytics (Umami)
Happy reading & hacking
Ozgur Gurcan
Software Engineer | Backend & Distributed Systems
Featured posts
Multi-Datacenter Active-Active Architecture: Close to 7M RPM on the Checkout Path
Architectural breakdown of running checkout-path services active-active across two datacenters at close to 7M RPM at peak: XDCR conflict resolution, Kafka mirroring, progressive zone shutdowns, and holding p99 at 63 ms on the main read path through load and chaos testing.
Zero-Downtime Database Migration: From Couchbase to PostgreSQL
How we migrated a critical production service from Couchbase to PostgreSQL with zero downtime, dual-write synchronization, and automated consistency verification.
Analyzing 32 TB of Coupon Data: Designing Dynamic Allocation to Eliminate System Waste
How querying a 32 TB coupon dataset in BigQuery showed coupons were distributed uniformly regardless of discount value, unmasked a silent country-level zero-coupon script bug, and led to a quality-based dynamic limit algorithm.
Taming the N+1 Query Problem at 4 Million Requests Per Minute
Redesigning our core domain data model: embedding related entities into single documents to remove an N+1 lookup costing up to 200 extra DB reads per request, retiring around 300M documents, 245 GB of RAM and 780 GB of disk at full rollout.
Recent posts
- Published on
A build-time NODE_ENV that broke my first deploy, a silent Node 18 fallback, curl-less health checks, stale image tags, metrics and disk blind spots: ten Coolify gotchas and their fixes.- Published on
With its MongoDB Atlas cluster gone, I moved JSONShare to a single Postgres jsonb table, upgraded Next.js 13.4 to 16, added Sentry, and self-hosted it on Coolify.- Published on
One PostgreSQL 18 server, one database and one role per project: the four statements I run for every new app, the defaults they fix, and how to prove role A can't open database B.