I build the part of Upstream nobody is supposed to notice. Servers, networking, storage, and the automation that keeps thousands of channels live without anyone sitting there watching them.
Before this I spent about ten years building backends for other people, usually leading a small team from an idea to something in production. Now I do it for a platform that is not allowed to go down.
What I do
- Live video pipeline
- Ingest, transcoding and delivery for streams that are never meant to stop. RTMP and WebRTC in, several platforms out, and the failure modes that only surface once a channel has been live for months without a restart.
- Backend architecture
- Twelve years of production backends. APIs, queues, data pipelines and the boring decisions that decide whether a system is still maintainable three years later.
- Infrastructure
- Bare metal and cloud, containers and orchestration, networking, storage and the backup policy behind it. I run the machines, not just the code on them.
- Scale and automation
- Systems designed to run unattended. Thousands of concurrent channels, and a platform that does not need somebody awake to keep it up.
- Building the team around it
- Hiring, process and the handover from idea to shipped. I have led small teams through platforms that grew past thirty thousand paying subscribers.
What Upstream is
Livestreaming software for people who would rather not run the machinery. Creators, radio stations, agencies and businesses stream through it every day. Four pillars:
- 24/7 live streaming
- Cloud channels that stay live around the clock, running playlists on a schedule with no machine of your own left switched on.
- Live Studio
- Invite guests, share screens and add branded overlays, all in the browser. Go live from a webcam, from OBS, or from video you recorded earlier.
- Multistreaming
- One live stream broadcast to several platforms at once: YouTube, Twitch, Kick, Facebook, TikTok, Rumble or any RTMP destination. Up to ten at a time, and it runs from the cloud so it costs your own machine nothing.
- Pre-recorded live
- Record on your own time, go live at the right one. The broadcast runs from Upstream’s servers rather than your computer, so you can be in the chat with your audience instead of operating the stream.
What I believe
Automation is most of the job. It is what lets a small team run infrastructure that never sleeps, and every manual step that disappears is time handed straight back to building the product.
Scale is not a milestone you pass once. Every time the platform grows, the tightest constraint moves somewhere new, so the job is finding it before it starts to matter. Some of that is thinking done early. Most of it is paying attention as things change.
Monitoring only counts if something acts on what it sees. Most of ours does. The platform watches itself, handles what it already knows how to handle, and pings me only when it genuinely needs a person. Anything less than that is a dashboard nobody is looking at.
The best infrastructure is the kind nobody ever mentions. When it works, people forget it is there and get to spend their attention on the thing they actually came to do. That is the whole point.