Built slowly.On purpose.
The work began behind 01:30 bakery shifts and finished after school days. Two years on, the discipline is what remains — the rest is production AWS infrastructure, AI-native tooling, and full-stack systems, designed on time horizons measured in years from a small desk in Adana.
Coordinates of practice.
Build window · open·Adana·GMT+3
- Field
- Adana · GMT+3
- Build
- Cloud Waste Hunter v2
- Reading
- Kleppmann · DDIA
- Stance
- Long arcs · daily practice
The system that builds the system matters more than the system that ships this quarter.
The discipline is the design.
Less a regimen than a rhythm. Calm repetition; same desk, same chair, same first hour. What gets shipped is the residue of what gets done quietly, day after day.
Most decisions here are made on a five-year horizon. The right system rarely ships this quarter. The wrong one always does.
The first work block lands before any notification does. Architecture decisions belong to the quietest part of the morning; the rest of the day owes them less.
Every production resource provisioned in Terraform. No console-clicked surprises. If it can't be re-created from a repo, it doesn't exist yet.
The same operating system runs both. Strength training in the early evening, code in the build window after. One discipline keeps the other honest.
Live builds.
Consistency over intensity.
Two years of mostly-daily commits. Some days are full ship-days; some are a single PR. Not every cyan square is a win — most are just showing up to the same desk and writing the next file.
How the work gets made.
Production-first.
Every system designed for real users from day one. No prototypes wearing the costume of products.
Infrastructure as code.
Reproducible. Version-controlled. Re-creatable from a git clone — or it does not count as deployed.
Cost-aware engineering.
Every architectural decision considers $/request. At scale, those choices compound faster than feature velocity.
AI as leverage.
Augment with AI; never replace engineering rigor. Models hallucinate. Types and tests don't.
Where the time goes.
Cloud Architecture
Production AWS infrastructure provisioned through Terraform. Multi-region patterns, IAM hardening, CUR-driven cost analytics, cross-account scanning via STS AssumeRole. From single-VPC SaaS to platforms that operate inside customer accounts.
AI Systems
LLM orchestration with Claude on AWS Bedrock. Streaming chat over Lambda Function URLs to bypass API Gateway timeouts. Master-prompt engineering for autonomous agents and structured remediation pipelines that ground model output in real data.
ClaudeAnthropic SDKAWS BedrockStreamingPrompt Engineering
Production SaaS
End-to-end products. Auth, billing, scanning engines, observability. Lemon Squeezy + Cognito + Sentry + PostHog. Customers see outcomes and dashboards; the infrastructure stays out of the way.
Lemon SqueezyAWS CognitoSentryPostHogFastAPINext.js
On the bench, this week.
- Building
- Cloud Waste Hunter v2 — cross-account scanner with Bedrock-grounded remediation pipelines.
- Reading
- Designing Data-Intensive Applications — Kleppmann. Slowly, and with notes.
- Operating from
- Adana, Türkiye · GMT+3. Coast roads and quiet rooms.
- Training
- Strength block, week four of an eight-week cycle. The deadlift is the gauge.
A walk before the keyboard sees a problem. Long stretches with no input. The day's most useful sentence is usually the one written down at the end of one of those walks — solitude isn't the goal, it's the operating condition.
Building toolsengineers actually use.
Smaller systems, sharper edges, fewer dashboards. Infrastructure that engineering teams can hold in one head. Operator time is the scarcest resource in the stack; the work here is built around protecting it.
On the hours that aren't code → /pulse
