Zoltan Petrik
Core skills forged over a decade of owning the outcome, not just the codebase.
Technical generalist — amplified by confident use of AI.
About
With 10+ years across full-stack web development, QA, sysadmin, and client-facing technical work, I've learned to see software as a whole system: the users, the business process, the implementation, the tests, the deployment, and what happens after release.
I care about the full lifecycle, from requirements and implementation to production, quality, and how the end result serves users and business needs.
These days, I'm focused on AI-augmented development and workflow automation, building SaaS products and practical tools that hold up in production.
Wore all the hats
End-to-end,
from idea to production.
Planned
Turned ideas into clear, workable plans that balanced business needs, scope, and technical reality.
Built
Translated plans into working software across the full tech stack; interface, logic, data, and workflows.
Tested
Made sure products worked as intended, including the edge cases and assumptions that never made it into specs.
Shipped
Took software into production, ensured deployment readiness, monitoring, and fast response when issues arose.
Supported
Stood behind products after launch: fixed issues, helped customers, and adapted as requirements evolved.
What I Bring
AI & Automation
AI-Augmented Development
- Full products built with AI-assisted workflows
- Claude Code, OpenAI Codex, OpenRouter API
- Prompt engineering & workflow design
- Agent orchestration
- MCP
Business
Working with People
- Requirements gathering & Stakeholder communication
- International client support (English)
- Freelance client work
- Ticketing; Jira, Monday, Redmine
Quality Assurance
Test Engineering
- ISTQB CTFL certified
- Coordinating testing across teams
- Complex systems (SAP, HubSpot, E-commerce)
- E2E Automation; Playwright, Cucumber
- Manual; UAT, Integration, Smoke, Exploratory, Mobile
Development
Full-Stack
- PHP, Node.js, TypeScript, React
- Mobile-Responsive HTML/CSS
- REST API development
- PostgreSQL, DB schema design
Infrastructure
Servers & Deployment
- Servers; Linux, Databases, Shell scripts
- Managed Cloud; DigitalOcean
- Networking; DNS, SSL, Apache, Nginx
- Deployment; Git, Server provisioning
- Production monitoring & on-call
Product
Idea to Working Software
- Built and shipped SaaS products from scratch
- UI/UX design in Figma
- Spec writing & architecture decisions
- User onboarding & documentation
- Analytics setup (GA/GTM, Tracking Pixels)
Experience Summary Download CV
2025–Present
Product Engineer
Independent Projects
What started as curiosity about AI tools quickly showed results and turned into conviction.
Evenings and weekends, I've been shipping useful projects end-to-end with AI-assisted workflows: design, build, deploy, iterate.
2022–2026
QA Lead
Fine & Rare Wines Ltd.
Hired to build E2E test automation; built in-house Playwright/Cucumber tooling and maintained production health monitoring throughout.
The initial UAT during a major CRM/ERP/catalogue migration gave deep exposure to every system and department: buying, sales, warehouse, CX.
That cross-system understanding became the basis for growing into QA Lead, owning quality across the full stack, from routine releases to major launches.
Managed 4 team members over the QA Lead period to increase testing throughput.
2020–2021
Sitebuilder
Webb & Flow Ltd.
Built high-quality, SEO-optimized sitebuilds from provided website designs while maintaining and extending the in-house frontend quality standards documentation.
Built custom in-house live-refresh sitebuild tooling that mirrored the production CMS's component structure, closing the gap between sitebuild handoff and CMS integration.
2016–2020
Co-Founder & Full Stack Developer
Asolve Kft.
Co-founded Asolve and built the initial SaaS product solo, a system to monitor and control render farms for 3D visualization studios.
In the early stages owned the product end-to-end: architecture, development, deployment, testing, monitoring, analytics, documentation, and customer support.
The prototype gained enough traction to enable a venture capital raise, which funded an expanded team and a second-generation product.
2012–2015
Freelance Webmaster & Marketing Assistant
Business Boost, Inc. + various clients
Ran web and marketing projects from brief to delivery for US-based clients, fully remote: building websites, landing pages, small web apps, and automations, plus PPC and analytics.
Available for hire
Full-Time
Remote or hybrid. Looking for a team where I can do great work, keep improving, and be proud of what we build.
Freelance / Contract
Available for SaaS products, internal tools, and automation workflows. I can handle projects from start to finish or help with specific development tasks.
Philosophy
For me, the work starts with
why it matters and who it's for.
Commercial Projects Shared with permission
SEOGalaxis
SEO Keyword Optimization SaaS for the Hungarian Market
Most Hungarian site owners don't know which keywords they're already ranking for but not optimized for. The data exists in Google Search Console, but surfacing actionable opportunities means tedious manual cross-referencing against page titles and headings. Existing tools are prohibitively expensive in this market segment.
Built with an SEO domain expert, the product automates this manual audit at scale: connect your Search Console, and within minutes see exactly which high-traffic keywords are missing from your titles, prioritized by search volume and opportunity score.
✅ Serving customers in production. Users published optimized content within hours of their first report.
Technical details
Designed and built end-to-end: React reporting interface, Express API with Auth0 multi-tenancy, a custom PostgreSQL job orchestration engine processing thousands of async jobs per report, and integrations with Google Search Console, DataForSEO, and a Hungarian proxy-based scraping pipeline.
Real-time Pushover alerting and a monitoring dashboard keep the service reliable for customers.
Amigo ERM
Sports Equipment Rental Management System
A Hungarian ski and snowboard rental shop was writing contracts by hand and tracking equipment availability on paper. During peak winter weekends, that meant longer queues, some double-booked gear, and a paper trail prone to missed entries and errors.
As a business-critical system, it had to run reliably through a full November-to-April season on shop-floor iPads with no room for failure; a crash in January could've caused serious headaches and reputational damage.
✅ The new system performed exceptionally: full season, zero incidents, zero unplanned downtime.
This first deployment doubled as a live beta. We're planning to offer it to other rental shops.
Technical details
I built the system covering the full rental workflow: customer intake, equipment fitting (binding angles, stance, etc.), real-time availability checks, and ready-to-sign printed contracts in Hungarian.
Multiple staff use the system at the same time. Equipment already booked for overlapping dates shows as unavailable in real time, so staff don't double-book by accident. Once a rental is signed, any further changes to its equipment, dates, or pricing are prevented on the database level.
Each customer gets their own fully isolated DigitalOcean VPS. Provisioning is scripted (VPS bootstrap, WireGuard mesh, Docker stack), repeatable in minutes.
Each VPS runs a Docker Compose stack (Caddy + Apache + PostgreSQL) gated entirely by WireGuard, with no public HTTP/HTTPS exposure. Database is Postgres with nightly encrypted off-site backups; schema changes ship as versioned migrations.
Sentry covers error tracking across backend and frontend for observability.
SurfersPoint
Custom React SSR Catalogue for Retail
A Hungarian surf shop needed a fast, current product catalogue that ranked well in search. They didn't need a webshop; customers come in to try gear before buying.
✅ Usability and on-page SEO were main priorities, achieving solid Mobile Lighthouse scores throughout:
Technical details
I built a custom React SSR catalogue with filtering, variant selection, and fluid responsive design. Every page is structured for SEO.
A 12-step automated data pipeline transforms raw wholesaler Excel into production-ready data: 2,245 rows from three suppliers normalized into 349 products with AI-generated Hungarian SEO copy, processed images, and human-in-the-loop overrides.
Next season, one pipeline run updates everything.
Open Source
Napirend
Local-First Task Manager and Time Tracker
I couldn't find a great task manager with built-in time tracking, so I built one. Every option I found either lacked time tracking or locked it behind an expensive subscription. Napirend is the tool I wanted but couldn't find: a daily-driver task manager where estimating effort, tracking time, and reviewing the difference are first-class features, not a premium add-on.
✅ Been using this as my daily driver for months now.
Technical details
First-class MCP integration: any MCP-compatible AI assistant (Claude, in my case) can create tasks, start timers, update status, and read state, with changes syncing to the UI in real time via SSE. Useful for hands-free task tracking while in flow, or for AI agents to track their own progress on long-running work.
Local-first by design. Data lives in a JSON file on your machine: no accounts, no cloud dependency, no subscription.
Daily Nutrition Planning Tool
Most people make eating decisions with no real-time feedback: no way to see whether today's meals actually meet their needs. KajChi closes that gap: you pick foods, the system instantly shows how you're doing across 140+ nutrients, personalized to your age, weight, and activity level.
I started with a more ambitious goal (fully automated meal planning), but learned that algorithms can't compose satisfying, edible meals. The better answer was to keep humans in the loop and give them a fast feedback instrument instead.
Technical details
Built offline-first with no backend, so nothing about your eating habits ever leaves your browser. All nutrient science is documented and every computed target can be overridden. The estimator is a starting point, not the final word.
Duma
Async Communication Layer Between AI Agents and Humans
AI agents sometimes need to ask the human a quick question. Instead of blocking until you respond, Duma lets them post the question, keep working on something else, and pick up the answer when it's ready. You answer with one tap or a voice note.
Used with my Claude Code setup: the agent posts a structured question via MCP; I get a macOS notification; I tap an answer or record a voice note (transcribed via Whisper); the agent polls for the answer when it's ready to use it.
Technical details
macOS menu bar app: Python + pywebview UI, SQLite storage, MCP HTTP server on localhost. The name means "talking" in Hungarian.