W3YZ Teknoloji.

Overview

W3YZ Teknoloji is a website-building platform similar to Squarespace or Wix, based in Istanbul. I joined as a freelancer with two months left on a six-month marketing deadline and 20+ mobile pages still unfinished. After on-time delivery I was offered full-time. When the senior lead left and recommended me as successor, I took over as de-facto tech lead with no documentation, no design-to-development sync, and three junior developers who needed guidance.

Problem

The platform had real customers and real traffic (20K+ daily requests) on a monolith that was not scaling. Designs and development drifted, the same UI patterns were being rebuilt in multiple places, and the AWS infrastructure had been set up by junior engineers in ways that were going to blow up eventually. Hosting costs were climbing for reasons nobody had measured.

Constraints

• Six-month marketing deadline with two months left and 20+ pages unfinished
• No internal documentation from the outgoing tech lead
• Three junior developers needing guidance
• 20K+ daily requests already live. Migrations had to happen without outage.
• AWS infra had risky single-server choices baked in. Untouchable in a deadline crunch, but had to be addressed before it broke.

Decisions

• Refactor the duplicated UI patterns into reusable components before building the remaining 20+ pages, instead of finishing them one at a time. Slower start, faster finish.
• Spend the first week of the lead role on handover sessions with the outgoing tech lead, writing English documentation in Notion, and drawing architecture diagrams in Miro. Junior developers cannot mentor themselves into a system that is not written down.
• Break the monolith into separate services: user management, template rendering, analytics. Dockerize deployments. Standardize AWS.
• Build a dedicated image compression service for user-uploaded media. The largest hosting cost was bandwidth, not compute.

What I built

• 20+ mobile pages with animations, dashboards, purchase flows, and analytics, delivered before the six-month marketing deadline
• Component-based UI architecture replacing duplicated patterns across the platform
• Comprehensive English documentation in Notion, architecture diagrams in Miro
• Mentoring for three junior developers across the API rewrite
• Monolith broken into microservices: user management, template rendering, analytics
• Dockerized deployments standardized across the AWS infrastructure
• Image compression service: 3.1MB images converted to ~156KB WebP (99% reduction)
• Infrastructure separation policy across the team after the Dubai Festival recovery

Hard problems worth naming

One week before a festival in Dubai, the AWS hard drive holding three VIP client websites (hoi.com.tr, dubaivizeal.com, prelogos.com) was deprecated. Junior developers had deployed all three sites and their databases on a single EC2 instance. All data was lost. I created a new Lightsail instance in the same AWS region, detached the damaged hard drive, attached it to the new server, downloaded everything locally, restored the databases, and cleaned and redeployed all three sites. Everything was live within days, before the festival, with zero data loss. The infrastructure separation rule (one application per server, databases off the application server) became permanent across the team after that.

Result

• 50% system efficiency improvement
• 45% monthly hosting cost reduction (~$10,800 annually)
• 99.9% uptime handling 20K+ daily requests
• 30% improvement in feature delivery speed after the design-development sync work
• 99% image compression (3.1MB to ~156KB WebP), reducing bandwidth cost and load times
• Three VIP client websites fully restored before a hard external deadline, zero data loss
• Three junior developers mentored through the API rewrite

Reflection

Most of the engineering work here was not heroic. It was infrastructure hygiene that should have happened earlier. The lesson I took: when you inherit a system without documentation, your first week is not for shipping. It is for capturing the mental model of the person who is leaving. Once that is on paper, the rest of the team can mentor themselves into competence. And when an AWS hard drive dies a week before a festival, do not panic-rebuild. Lightsail in the same region, attach the dead drive, download, restore, redeploy.