Skip to content
View brjen's full-sized avatar

Block or report brjen

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
brjen/README.md

Hi, I'm Brian 👋

I build and operate self-governing AI systems — and I run a production business on top of them.

Self-taught, outcomes-first. Over the past few months I designed and built Gridline, a private multi-agent AI operating system: AI agents that work under a shared contract, remember across sessions, review each other's output, measure their own performance, and improve over time. The AI writes the syntax — I architect the system, the governance, and the operating discipline.


🧠 Gridline — a self-governing AI operating system

Not a chatbot or a pile of scripts — a working AI environment built for sustained work at scale.

  • ~49,000 image renders through a fully tracked production pipeline (~137,000 artifacts under hash-based chain of custody)
  • ~1,700 work tickets executed (~1,337 completed) — every decision, reason, and outcome recorded
  • ~80 codified operating rules · ~18 telemetry databases · multiple coordinated machines
  • Multi-model agent orchestration with independent review gates before changes land
  • Measured reliability & self-improvement — 96% success across 1,300+ tracked tasks (zero hard failures); self-audit policy-violation rate cut ~12× (2.3% → 0.18%) even as activity tripled
  • Custom render engine (5 generations) that replaced commercial tools — tuned to run oversized models on constrained hardware

Work → record → review → measure → learn → update the system. That loop is the point.


🛠️ Infrastructure — bare metal, self-hosted

A multi-machine estate I run end to end:

  • Compute: AMD RX 7800 XT · NVIDIA GTX 1080 Ti · Apple Silicon
  • Virtualization: Proxmox VE
  • Network: OPNsense · Unbound recursive DNS · AdGuard Home
  • Ops: Docker · systemd · Linux servers · RAID · disk encryption · full observability

Philosophy: own the complete stack, no vendor lock-in. I debug down to the metal — including a PyTorch memory fix derived from production profiling.


🧰 Stack

Python Node.js Svelte FastAPI SQLite Linux Proxmox Git


📌 Open source

pytorch-memory-fix — environment variables that eliminate PyTorch memory creep with zero code changes, derived from real production profiling.


📫 Open to work

Applied-AI · AI / agent engineering · forward-deployed & solutions engineering · consulting on AI operating systems and agent governance.

Portfolio · LinkedIn · GitHub

Popular repositories Loading

  1. pytorch-memory-fix pytorch-memory-fix Public

    Two environment variables that fix PyTorch/glibc memory creep on Linux. Zero code changes. Zero performance cost.

    Python 10 2

  2. brjen brjen Public

    Config files for my GitHub profile.

  3. brjen.github.io brjen.github.io Public

    Personal portfolio — AI systems engineer

    HTML