Hey — I'm Norbert
I build browser-native SaaS apps
I build SaaS on one stack, front-end heavy - and I push the browser past where it's supposed to go.
I've done it as a founder, as employee #5 and then #2 - the person a startup brings in early to build. Everything with javascript, top to bottom - TypeScript on both ends:
AI-assisted video editor in the browser, 60 frames per second rendered to MP4 with javascript
Process mining with javascript, client and server side. Extract, Transform, Load then Process. Data heavy, complex visualizations
Mostly browser-native SaaS
My free contribution to the world
A side project that turned out useful - and stayed alive instead of gathering dust in a drawer. Draw a truss structure on a canvas and watch the forces resolve in real time: no backend, no install. Used by students, teachers, engineers, and researchers all over the world, more than 10k visits a month, fully organic.
Early-stage Hungarian startup. An AI video editor that turns screen recordings into product demo videos.
I was the second engineer (first full-stack), and I owned the editing and rendering stack. Real-time editing at 60 fps in the browser, exported to MP4.
The rendering was decided on PixiJS before me. I kept building it, but PixiJS became a dead end, so I migrated everything to Three.js / React Three Fiber. I also built the slide editor with WebGL animations and custom shaders, and a custom layout engine.
I wrote all the algorithms for eg: camera movement, auto-zoom, cursor smoothing, transitions, auto-cuts. Captions. Thumbnail generation and publish to Webflow.
AI features: a narration pipeline (speech-to-text on the recording, an LLM writes the script, ElevenLabs generates the voice), and document-from-video (turns a recording into a structured text document).
Backend: auth, access control, asset management, rendering pipeline, cloud functions
US startup with Hungarian dev center, 70+ employee. React / Next.js lead - built an interactive plant-graph flowchart and a D3 + Canvas equipment analyzer, and rewrote the site-mapping tool from OpenLayers to the Google Maps API.
Process mining at scale, in the browser and server up to 100M+ events.
I joined a US startup in 2016 as employee #5. ABBYY acquired us in 2019, and I left when the company had grown past 1000 people. 6.5 years in total, 2.5 of them as team lead. One product, full lifecycle. I learned most of what I know here - too much to list.
Featured: pushing the browser's data ceiling from 10M to 100M+ events, running a React-in-AngularJS migration while the product kept shipping, and cutting a 100M-event calculation from ten minutes to a few seconds. ETL module to load data to Timeline in the proper format. Plus the team work itself - architecture, reviews, hiring, mentoring.
My first startup - a platform for engineers to build and share calculations.
I built a team, caught Autodesk's attention, won several competitions, and turned down a term sheet. It didn't make it - but it's the reason I write code today. Failing a startup has one upside: it made me a 2× Fuckup Nights speaker (2019 & 2020)
Small engineering firms (Mérnök-Mátrix, FLS) and a global brick multinational (Wienerberger), in and around steel and concrete. ~70 structural projects, structural analysis, drafting and steel detailing, energy-efficient-masonry R&D and 100+ trainings a year as Wienerberger's technical consultant, covering half the country.
I decided to become an engineer at 12 and never really changed my mind - I just changed the material.
Before code, there was pasta: I'm a world champion in spaghetti bridge building - yes, that's a real thing - though it was a while ago now (Canada, 2008 & 2009). I started working in those same years, then finished the degree and spent 6 years in construction, around steel and concrete.
Then Exelean - my own startup. It didn't last, but one lesson stuck: I'd rather build the product than run the company. So that's what I've done ever since - joined startups early, as a key engineer, and owned the parts that are hard to get right.
Since 2015 it's been software, and the work is the same shape: the hard performance problems, the edge of what a browser can do - 100M events in a tab, 60 fps editing, things that aren't meant to run client-side. I want to know where a system breaks, and how far I can push it first.
I build AI-augmented - Claude Code today, after Junie and Cursor. I switch when necessary.
Outside the browser: I'm a father, I run, fix things and I'm easily talked into a walk in the forest.
Happy to talk - and not just tech. Structures, minimalism, parenting. Reach out about a role, or about none of the above.
Open to work
Currently open to senior, staff, team-lead, or founding-engineer roles - at startups and established product teams alike. Budapest (hybrid), or remote across the EU and US. I do my best work on small teams where one engineer's range actually matters - but I'm open to the conversation either way.
