Discuss a project Software architecture · AI/ML · full-stack delivery
I'm Onur Yilmaz—an independent software architect and hands-on engineer. I design and build production systems, AI/ML products, full-stack platforms, and the difficult architecture between an ambitious idea and a reliable release.
01Architecture
that survives reality
02AI/ML
inside real products
03Hands-on
production delivery
01 One accountable senior builder
02 Architecture and implementation together
03 Technology chosen for the problem
Career references
Since 2012, I've architected and shipped products across global brands, regulated industries, industrial operations, mobility, and AI-native startups.
Klarian
Mapilio
Brawn
Himalaya Pay
AvivaSA
Honda
Bupa Acıbadem
Unilever AlgidaCompany and product marks are shown solely as factual career and project references. Engagement types include employment and independent contracts.
What I build
I'm most useful when a product is technically demanding, the path is unclear, and you need someone who can make the decisions and deliver the software.
I turn unclear, high-stakes requirements into software that can actually be built, operated, and extended. System boundaries, APIs, domain models, security, performance, integration, and delivery strategy—thought through together.
I build AI into real products: custom agents, LLM and model integrations, fine-tuning, MCP tools, knowledge-grounded workflows, computer vision, and the surrounding software that makes them dependable.
Hands-on delivery from interface to infrastructure. I build web and mobile products, backend services, production APIs, cloud environments, and automated delivery pipelines—while improving how the team discovers, builds, tests, reviews, documents, and releases software with modern AI coding agents and tools.
When software depends on fragmented operational, spatial, relational, or graph data, I make that foundation usable—without letting the data layer become the product story.
The differentiator
I start with the operating reality—not a favourite framework or an inherited answer. When a requirement calls for offline behaviour, native capabilities, cross-platform reach, legacy integration, or an unusual delivery model, I combine proven technologies in ways conventional stack thinking can miss.
The architecture can be original. The result still has to be maintainable, explainable, and ready to run.
AI product engineering
Useful AI needs product thinking, software architecture, evaluation, reliable tools, domain context, guardrails, observability, and a way to fail safely. I build the complete system—not a clever demo that stops at the API response.
From custom agents and MCP infrastructure to fine-tuned models, knowledge-grounded applications and computer vision, I choose the approach that earns its place in production.
I also apply the latest coding agents and AI engineering tools across the complete software lifecycle—from discovery and architecture through implementation, testing, review, documentation, release, and operations. The aim is better judgement and faster feedback, not automation theatre.
Discover & architectRequirements, options, boundaries, risk
Build & verifyCoding agents, tests, review, documentation
Release & operateAutomation, observability, control, iteration
What I've built
Real products from recent AI platforms to more than a decade of mobile, industrial, financial, and enterprise software. Klarian references are independent contractor work, not employment or an implied partnership.
Full-stack platform engineering across product, APIs, cloud infrastructure, knowledge graphs, MCP, and custom agents—turning complex operational context into usable software.
Application architecture and hands-on engineering for a sophisticated geohazard platform, including an unconventional cross-platform approach that brings rich mapping, native capabilities, and reliable offline field workflows together.
Full-stack development and performance optimisation for a street-image collection platform operating across more than 20 countries, spanning mobile capture, web product, backend, and ML workloads.
Built for WhatsAppBuilt the software behind a white-labelled WhatsApp member journey: goal-based onboarding, personalised training and nutrition plans, coach matching, and automated follow-up—without asking members to install another app.
A Django and React product combining fine-tuned GPT models with authenticated large-scale web automation, custom model workflows, and production-grade application engineering.
Defined the greenfield solution architecture, coding and integration standards, and led native mobile delivery across iOS, Android, middleware, microservices, security, testing, and release.
Architected mobile and middleware foundations, established digital products, and led the CRM roadmap to more than 1,200 daily users and an 800% rise in platform usage—connecting technical delivery with measurable adoption.
Designed and developed the Android client, administration application, and middleware as one exhibition system showcasing the Galaxy Note pen experience.
Designed the iOS client and middleware for production, planning, and sales data, including custom graph interfaces and support for Japanese and European calendars.
Designed native iOS libraries and used them across policy-management and internal applications, creating a reusable base instead of solving each mobile product separately.
Architected and built a rich tablet system for fault and scheduled maintenance management, connecting visual factory-floor workflows with enterprise systems.
Built the native iOS reading experience for TTNET’s e-book platform, integrating the Adobe Reader mobile SDK with Apple’s application framework.
How I work
FLIKE is my consultancy. You work directly with me: one experienced software architect who can move from a boardroom constraint to a system design, then into the codebase to make it real.
I can own a difficult technical stream, accelerate an existing engineering team, rescue a product that has lost direction, or take a greenfield idea from architecture to production.
I'm stack-agnostic because architecture is about trade-offs, not allegiance. I also improve the development lifecycle itself, integrating AI coding agents and modern engineering tools where they make delivery faster, safer, and easier to sustain.
Tools & platforms
Sector experience
Hard problem welcome
If the software is complex, important, or still undefined, send me the honest version. I'll tell you how I can move it forward—and if I'm not the right fit, I'll say so.
onur@flike.co.uk