Open to Full Stack, Frontend, Backend, AI and QA Engineer roles
I write TypeScript from the browser to the API to the database, and I own what happens between them.
5+ years of TypeScript, React and Node.js on products people use every day, plus LLM pipelines running in production.
Frontend
Backend
AI
On my last platform there were two core developers, so every layer was ours to get right. These habits come from that, and from starting out in QA.
Schema, API, state and UI, plus the client conversation before them. On a two-person core team there was no one to hand a layer to.
When the same bug keeps surfacing in the UI, the cause is often in how the data is stored. Changing that is usually cheaper than rewriting the code around it.
Two years of testing came before my first developer job: 100+ defects logged and verified across every major browser engine. That’s where I learned how edge cases and API contracts fail.
At Chanalyze I met enterprise client stakeholders every week to work through technical and UX questions with them directly.
Developer documentation and company-wide guides for the design system I maintained, plus three published articles and papers.
Coding agents are part of my daily workflow, from test harnesses to a first pass of code review. How I use AI
Four roles since March 2019, newest first. It started in QA, which is where I learned how software actually breaks.
One of two core developers on an enterprise analytics platform for medical research. I owned the database architecture, the backend APIs and the client-side state, and met enterprise client stakeholders every week to advise on technical and UX questions.
Technologies
Frontend work for Web3 platforms, NFT minting and browser games, with real-time state management throughout.
Technologies
Enterprise tourism management platforms: backend REST APIs and dense data-grid interfaces.
Technologies
Led quality assurance and partner integration support for the company’s core ERP system, with a focus on reliability, API contracts and edge cases.
Technologies
I’ve shipped LLM pipelines to production, built my own RAG system, and work alongside coding agents every day.
In production, at Chanalyze
A production Azure OpenAI pipeline that pulls and filters relevant research documents, builds dynamic prompts, monitors token counts and spendings, handles retries, and validates all structured outputs against JSON schemas.
Every day
Claude Code, OpenCode and Google Antigravity are part of how I build. I set up the agent environment for each project and review what the agents produce.
What I have been working on since leaving Chanalyze: two client websites taken from concept to launch, and a RAG system built to learn vector search hands-on.
Client work, 2026 · live
Design and UI architecture for a textile embroidery and garment printing company in Stuttgart (einfach-stickerei.de). Clean, conversion-focused layout showcasing corporate fashion, custom embroidery services, and quote inquiry flows.
Client work, 2026 · live
A landing and services site for a physiotherapy centre in Labin. I did all of it: client discovery, UI design, the build, and deployment on Vercel with a custom domain.
Self-directed, since June 2026
Retrieval-augmented generation on PostgreSQL with pgvector, built to learn vector search hands-on, end to end.
Grouped the way most job specs ask for them.
Filled: used in my latest role (Chanalyze, 2022 to 2026) or every day now.
2024
Tria d.o.o.
Focus Responsive web architecture, performance-first mobile design, Core Web Vitals, and search visibility.
2023
Tria d.o.o.
Focus State persistence, a CSS variables architecture, accessibility (WCAG), and transitions between themes without a flash.
This page’s theme switch is built the same way.
2020
PRINTING & DESIGN 2020 Conference, peer-reviewed, co-author
Focus GIS data processing, cloud infrastructure scalability, and spatial information systems.
Faculty of Informatics, Juraj Dobrila University of Pula
Faculty of Informatics, Juraj Dobrila University of Pula
Case 01 · Healthcare analytics, interactive dashboards
When User A filtered a shared dashboard, User B’s screen refreshed with User A’s filters. I fixed it in days by changing where filters are stored.
An enterprise analytics platform let teams share interactive dashboards. Every filter change wrote straight to the shared report entity in the database, so anyone else looking at that report got someone else's filters. That cross-user bleed had been open for 4 years and was blocking several roadmap deliverables. Senior leadership had already assessed a full-stack decoupling refactor and judged it high-risk and too costly for a system in active use.
I left most of the application code alone and solved it in the data model, with a User-Scoped Configuration Overlay.
user_report_configs, stores each user's filter adjustments strictly as state deltas.Simulation, sample data
Pick a period filter as User A or User B. With a shared report entity, the other screen changes too. Switch to the overlay and each filter lands in user_report_configs instead.
Database
| id | filters |
|---|---|
| shared_report | {"period":"All"} |
| user_id | report_id | delta |
|---|---|---|
| 0 rows | ||
Both users are looking at the same shared report.
Architectural lesson
Changing the data model is often safer and faster than refactoring the code around it.
Multi-tenancy, a design system, an LLM pipeline, and real-time Web3 and 3D. Each one lists the stack, the decision and what changed.
Case 02 · Medical research platform, enterprise SaaS
Multi-tenant isolation in PostgreSQL and MongoDB, so competing clients share a platform but never each other’s data.
The platform served over 10 enterprise clients, each on its own dedicated configuration and instance. Infrastructure overhead was heavy, hosting bills were high, and every deployment was complex and easy to get wrong.
I designed group-based tenant isolation for both the relational and the document database, then moved every client environment into one scalable deployment.
Case 03 · Enterprise ERP, CRM and custom client apps
A shared React and TypeScript design system of 30+ components, maintained independently.
Teams building ERP and CRM client modules kept duplicating code, the UI drifted from product to product, and each new project began with slow setup.
I was the main maintainer of the internal component library, published as private npm packages, and built strongly typed, accessible, reusable components with strict props contracts. Theme switching, dark mode and layout standards became shared by every app that used it.
Case 04 · Biotech, medical research data extraction
A production LLM pipeline with dynamic prompt assembly, token tracking, request retries and schema-valid output.
Researchers spent hundreds of manual hours reviewing dense, non-standardized medical research documents and pulling structured parameters out by hand.
I built a dedicated Azure OpenAI integration in Node.js and TypeScript. The backend pulls, filters and sorts relevant research documents, constructs dynamic few-shot prompts, and manages requests with automated retries. Every step is tracked through PostgreSQL, auditing the pipeline lifecycle, token counts and API spendings, while strict JSON schema validation ensures only clean, structured data reaches the database and surfaces in interactive React and AnyChart dashboards.
Case 05 · Web3 gaming and asset minting
Three separate pieces of frontend work at Valere Margins, each with its own hard constraint.
The complete presentation and transaction layer for an Ethereum asset system: MetaMask wallet sign-in, Ethers.js contract calls and backend transaction tracking.
Real-time, two-way state sync over WebSockets for a web RPG that reads Solana blockchain state, with optimistic UI updates to keep it responsive.
I profiled a Three.js scene, then cut draw calls, optimized geometry and lazy-loaded assets. Frame rates and page performance improved by 50%.
Email is quickest. Tell me about the team, the stack and the system that needs to change.
I’m in Čakovec/Varaždin, Croatia, on Central European Time, and open to remote or hybrid roles in the EU and worldwide. As an EU citizen, I don’t need sponsorship to work anywhere in the EU.