k.bKubilay
Bayraktar.
/
Let’s talk

SOFTWARE ARCHITECT · TEAM LEADER · AI PRODUCT BUILDER

Engineering systems.
Building possibilities.

I’m Kubilay. I turn complex problems into reliable software — and ambitious ideas into AI-powered products.

Explore my work Download résumé

BASED IN ISTANBUL · THINKING GLOBALLY

[ KB / SYSTEMS THINKING ]
ARCHITECTURE
INTELLIGENCE
HUMAN INTENT
+IDEA → ARCHITECTURE → PRODUCT+
19+

years in software

15

personal projects

02

AI products in development

Experience

EXPERIENCE ACROSS ENTERPRISE, CLOUD & INDUSTRY

metusTurkcellOBASEPIRELLIIntronis

01 / SELECTED WORK

Ideas, built into products.

Two independent projects. A shared belief: AI becomes useful when it is grounded in evidence and built around people.

ONGOING PERSONAL PROJECT
● ● ●agentwebtest.com↗
AgentWebTest public website showing AI agent website readiness
WEBSITEAI AGENTEVIDENCE

AI AGENT EVALUATION & WEBSITE INTELLIGENCE

AgentWebTest01

The next visitor is an agent.

AgentWebTest measures whether AI systems can discover, understand and navigate a public website. Deterministic audits and bounded AI agents turn inspected evidence into actionable findings.

THE ENGINEERING DECISION

Let code establish the facts. Let AI interpret verified evidence. Validate its output before it reaches the report.

Next.jsTypeScriptOpenAICloudflareSupabaseTrigger.dev
Visit project ↗
Product & purpose · How the agent works +

MY FOCUS

Product architecture, agent workflows, evidence verification and cloud delivery.

Product & purpose

Website readiness, recurring monitoring, regression detection and evidence-linked findings. Workspace roles, audit credits and subscriptions support the product workflow.

How the agent works

A deterministic crawl inspects a bounded set of public pages. The OpenAI Research Agent prioritizes verified findings; Agent Tasks evaluate site-specific questions. URLs and quoted evidence are checked against inspected sources.

Infrastructure & delivery

Next.js and React run on Cloudflare Workers through OpenNext. Supabase PostgreSQL and Auth manage tenant data and identity; Drizzle defines the schema. Trigger.dev handles durable jobs. Paddle and Resend support billing and notifications. GitHub Actions separates test and production.

ONGOING PERSONAL PROJECT
● ● ●muturio.com↗
Muturio public website demonstrating a personal AI agent
INTENTAI AGENTHUMAN CHOICE

PERSONAL AI AGENTS & SOCIAL DISCOVERY

Muturio02

AI prepares. People decide.

Muturio is a social discovery product built around personal goals. A personal agent helps people discover resources, prepare content and find relevant connections — with the user choosing the next step.

THE ENGINEERING DECISION

Use embeddings to understand intent, deterministic rules to rank mutual relevance, and explicit human decisions to make connections.

ExpoReact NativeHonoOpenAIpgvectorCloudflare
Visit project ↗
Product & purpose · How the agent works +

MY FOCUS

Universal product architecture, bilateral matching, AI integration and user-controlled workflows.

Product & purpose

Personal goals guide resource discovery, daily briefings, content drafts and people opportunities. The product spans web and mobile, with Android distributed through a Trusted Web Activity package.

How the agent works

OpenAI embeddings represent intent in pgvector. A bilateral scoring engine evaluates relevance in both directions; content drafting uses bounded model calls. Private intent stays on the server. Publishing and introductions require explicit user decisions.

Infrastructure & delivery

Expo Router and React Native share the product client across platforms. Static web runs on Cloudflare Pages; a Hono API runs on Workers. Supabase provides PostgreSQL, pgvector, Auth and RLS. Worker cron jobs process queued work; Resend handles transactional email.

02 / UNDER THE HOOD

Good AI needs good architecture.

Explore the systems behind the products. Select a component to see why it is there.

SELECT A COMPONENT ↘
01

01.

Evidence before inference.

Verified sources and structured outputs make AI findings inspectable.

02.

People stay in control.

Agents prepare possibilities; users decide what to publish or act on.

03.

Reliability is designed in.

Tenant boundaries, idempotent jobs and isolated environments support dependable delivery.

[ AGENT WORKFLOW / EXPLAINED ]

Follow the agent’s reasoning path.

An illustrative walkthrough of the AgentWebTest pipeline. No live scan or model call is performed.

  1. 01
    Inspect public pages
  2. 02
    Collect deterministic findings
  3. 03
    Interpret verified evidence
  4. 04
    Validate URLs & citations

Ready to explore

03 / EXPERIENCE

A builder’s mindset.
An architect’s perspective.

From industrial systems to cloud applications, I connect hands-on engineering with team leadership.

Download résumé ↓
2015 — PRESENT

Software Architect · Metus

Architecting and evolving enterprise applications with .NET Core and React on Azure. Building event-driven services, intelligent search and a culture of maintainable, tested software.

  • Microservices with Ocelot, Azure Service Bus and Redis; object storage with Amazon S3.
  • Elasticsearch integrated with a dedicated Python AI service for intelligent search.
  • 600+ unit, integration and in-memory integration tests using xUnit.
  • SOLID, CQRS and MediatR; ongoing refactoring and team workflow leadership.
.NET CoreReactAzurePythonElasticsearchxUnit

Earlier chapters

2013 — 2014

Turkcell

Product Development Specialist · Consultant

+

Developed a web-based budget planning platform with automated financial suggestions and inter-budget transfers.

2012 — 2014

Obase Software

Product Technical Manager

+

Built document and label design software, asynchronous ETL tooling, and a real-time franchising simulation engine for Carrefour.

2010 — 2012

Plano Engineering

Software Developer

+

Built CSV-to-KML geographic visualization tools and multithreaded utilities for high-volume telecom log processing.

2009 — 2010

Graphtec Interactive

Web Application Developer · Palm Desert, USA

+

Delivered frontend implementations at a company serving more than 300 websites.

2009

Intronis Technologies

Scrum Developer · Boston, USA

+

Helped migrate a backup and disaster recovery product from desktop to cloud. Worked with an Agile team across five countries and received direct Scrum coaching from Jeff Sutherland.

2007 — 2009

Pirelli Tyres

Software Developer

+

Contributed to the Motorsport/Rally factory IT deployment with the German core team. Developed touch-panel maintenance systems and mobile barcode traceability software.

2006 — 2007

Kocaeli Health Directorate

Software Developer

+

Developed hardware asset tracking and IT help desk applications, receiving recognition and an award from the Ministry of Health.

04 / THE PERSON BEHIND THE CODE

Logic meets imagination.

Software engineering is my profession. Curiosity is the thread running through everything else. Oil and stone painting, cooking and travel keep me looking at problems from new angles.

I enjoy turning ideas into working things — in a brainstorming session, in a codebase, or on a canvas.

↗

Industrial Engineering · Sakarya University

✳

Painting · Cooking · Travel · Basketball

[ BUILDER’S ARCHIVE ]

A history of shipping.

Beyond my current products, I’ve published reusable .NET layers, developer utilities, a physics engine and web applications on CodeCanyon.

01.NET Core Business Layer ↗02C# Simple Physics Game Engine ↗03SQL Builder & Organizer ↗04MVC Amazon S3 Uploader ↗

05 / WHAT’S NEXT

Let’s build something
that matters.

Looking for a software architect who leads teams, understands the full system and builds AI products hands-on? Let’s start a conversation.

Get in touch ↗
bayraktarkubilay@hotmail.com ↗LinkedIn ↗GitHub ↗Download résumé ↓