Engineering student & production engineer in the making.
I build the stuff that keeps live streams alive and audio clean.
- 🎥 ~2.5 years in video / live streaming: ffmpeg pipelines, GPU transcoding, archiving and validation services
- 🎚️ Live stream infrastructure: piplines, obs, analyze logs and record
- 🛠️ Almost 3 years in IT overall, from infrastructure to full-stack tooling
- ⚙️ I care about how things work under the hood, from ffmpeg flags to GPU pipelines, and use AI to speed up the boring parts
Roughly in order of how well I know them:
Skills & tools
- notGT: broadcast titles pack for OBS (NodeCG fork), corporate style, fullscreen graphics
- Streaming platform: Next.js + Prisma + PostgreSQL site for live streams, with TS microservices for archiving and validation, all in Docker Compose
- GPU transcoding node: ffmpeg + NVENC in Docker, GPU monitoring with dcgm-exporter
- vMixRTC: Customizable controller for vMix (reworked vMixUTC, but on Rust)
- companion-module-aimp-audioplayer: Bitfocus Companion module to control AIMP
- AIMP-HTTP-Remote-Control-Server: AIMP plugin with an HTTP remote API
- Homelab: WireGuard, Mihomo, Traefik, reverse proxy and DNS
- Budget local LLM inference: GPU selection and server build for running models locally
- 🎛️ Embedded StreamDeck-style controller with Bitfocus Companion integration
- 💡 DMX lighting control app on Tauri + Rust
- 📡 Serverless messenger for Meshtastic / LoRa mesh networks, plus an efficient compression scheme for sending larger data over LoRa
- 🧑🎤 Video-call codec idea: face capture to blendshape JSON + voice, rebuilt as an avatar on the receiver's side (Bulks)
- 🧠 Context optimizer for AI coding agents (Claude / OpenClaude): a cheaper model scans the project first and condenses it, then only the relevant context goes to Claude, cutting token costs
- 🌋 Learning Vulkan
