Özgür Şahin Çakmak

AI-native engineering

Output per engineer, per day

1. Day onebefore agents
2. 4.8×2026 average
3. 9.6×last quarter

• −68%cost per request
• −65%context per request

Ten agents run delivery — one product owner dispatching nine repository-scoped developers that code, lint, scan and test — with a named person approving every gate, on an in-house harness where agents are defined as data, so a new one ships without a release. Same three engineers throughout; changes reach production in nine days instead of thirty-four.

People & teams

50+Engineers led and mentored

• 1 → 11engineering org built
• 21cross-functional
• 10+nationalities

Across five companies, from a first five-person squad to an engineering organization built from a single engineer — and the structure people work inside as much as the headcount: an SDLC culture introduced across 23 repositories at one company, ad-hoc request handling replaced with a defined process at another.

Cost & efficiency

50%+Infrastructure cost removed

A monolith rebuilt as cloud-native microservices on AWS — half the infrastructure bill removed while scalability and regional deployment improved, rather than traded away for the saving. The same ground-up AWS work stood up the DevOps environment for a first international program elsewhere: the cost result came out of architecture decisions, not a procurement negotiation.

Maintenance & delivery

80–99%Fewer production issues

• +40–50%delivery capacity, same headcount
• 2defects closed after a year unfixed

On a bank's core Corporate Loans system and an insurer's core platform, both carrying production load for seven internal departments and their agencies. The gain came from replacing ad-hoc request handling with a defined process — the same change that lifted delivery capacity without adding a single engineer.

Transaction scale

Millionsof daily transactions handled

• 1M+per day, by design
• 7departments and their agencies

High-volume systems architected end to end: an e-invoice platform whose database was designed for more than a million transactions a day, years before "cloud-scale" was a common expectation, and an insurer's core carrying millions a day in production. Earlier still, a three-way real-time sync between an ERP, a B2B platform and a storefront — a country first at the time.

Program delivery

$4MTransformation program led

• 162services aggregated
• 40+KPIs
• 19engineers, two sub-projects

A PEGA rebuild of a bank's Corporate Banking platform, run across two sub-projects after being promoted into it from leading the five-person team on the bank's Corporate Loans core — so the decision system at its center was architected by someone who already knew what the existing platform did and where it failed.

Security & governance

SOC 2Type II program led

• 100%security policies approved
• 0critical or high findings open

Security controls, governance, policy frameworks, risk management and audit readiness owned end to end, on a platform handling enterprise customer data. An independent penetration test closed with every high and medium finding remediated, and disaster recovery is rehearsed to a verified full-platform rebuild rather than written down and filed.

R&D & funding

₺5MAI R&D program secured

• €2MEIC initiative in progress
• 2EU research projects delivered

TÜBİTAK-funded and still led today. Not a first pass at public research funding either — two EU framework projects, PaaSPort and IMAGINE Future Factories, were delivered years earlier, including a semantic search engine taken solo from an MSc thesis to a production system.