I own products end to end: architecture, code, infrastructure and operations. By day I lead engineering at a Swedish SaaS for newsletters, SMS and marketing automation, used by about 1,500 organizations.
The three projects below are my own: nights, weekends, my own time. Each one has a demo of concept you can open right here and, where the real tool is public, a link to it.
01 / 03
TurfGPS
2024 →in development
Turf is a GPS game played around the world: players capture zones by visiting them in person.
A route planner for a trip you are already making: say what you value and how many minutes you can spare, and it finds the most valuable route through the zones on the way.
how it works
01Enter the trip you are already making, from A to B, with stops if you like.
02Say how many extra minutes you can spare and which zone types you value most.
03It compares the roads and picks the stops that give you the most for those minutes.
Paper trading: the strategies run on live market data, but every order goes to a simulator, never real capital.
Autonomous algorithmic trading platform, designed, built and operated by exactly one person. It paper-trades only, while the strategies are tuned.
how it works
01Market data streams in through one adapter per provider; the engine never sees a vendor.
02Strategies score about 50 assets around the clock and propose orders.
03A risk gate checks every order before it reaches the paper simulator.
~50assets evaluated in real time
24/7autonomous operation
1engineer, across the entire stack
Go · React · TypeScript · PostgreSQL · WebSockets · GitHub Actions · Linux
closed source, ask me about it
preview · open the demo to use it
03 / 03
PepegaClapBot
2019–2024retired with honors
Twitch chat never sleeps: it keeps running whether the stream is live or not.
Livestream automation platform that grew from a hobby script into a small SaaS used by Twitch streamers.
how it works
01The bot joins each streamer's chat and reads every message.
02Moderation rules and engagement actions fire through an MQTT event pipeline.
03Streamers set it up once; it ran for five years across about ten channels.
170M+chat messages processedcounted by the bot itself: one increment per message read.
230K+unique chattersone increment per chatter the bot had not seen before.
≈23 yrsof chat monitoredsummed connect time across every channel. Twitch chat keeps running when the stream is off, so the bot listens around the clock.
works on my machine™✦0 critical bugs (source: trust me)✦170M+ messages later✦deployed on a Friday once, survived✦≈23 years of chat, monitored✦CSS is a programming language (fight me)✦reads the logs before guessing✦no cookies, only 🍪 jokes✦works on my machine™✦0 critical bugs (source: trust me)✦170M+ messages later✦deployed on a Friday once, survived✦≈23 years of chat, monitored✦CSS is a programming language (fight me)✦reads the logs before guessing✦no cookies, only 🍪 jokes✦
projects
Smaller builds, experiments and university work
🍅24
Is My Tomato Sick?
AI-powered disease detection for tomato leaves. A real product. Real tomatoes were saved.
