>cd /home/thushan

IoT, AI, Rust, Go, TinyGo

© 2026 Thushan Fernando CC BY-SA 4.0
Built with Hugo ❤️ hyde-hyde and forked.

About Thushan Fernando

Thushan

I’m a Senior Consultant at SixPivot (formerly Readify) and Founder of TensorFoundry based in Australia.

TensorFoundry started in the 2010s as something I’d tinker on after hours and officially became a business on April 1st, 2025. The date is genuine, the timing was not a joke - but also, what kind of fool would start an AI company? TensorFoundry is where I take on NVIDIA consulting and work with customers on edge AI, fine-tuning and inference challenges. It’s also where the tools live (yes, TF fits my initials. That wasn’t on purpose):

  • Olla - smart load balancer and proxy for LLM infrastructure (open source)
  • FoundryOS - distributed GPU cluster management for self-hosted inference (early adopters)
  • Forge - pure-Rust single-binary inference server for NVIDIA GPUs (active development)
  • Alloy - LLM gateway and control plane for managing multiple providers (approaching v1)
  • Kaizen - terminal AI coding agent with persistent memory (public, v1)
  • Pivotal - self-hosted knowledge platform with AI agents that build and maintain a searchable wiki from your documents (internal)

The common thread is letting you run AI workloads on your own infrastructure instead of paying for managed services.

When digging into the why and how behind all this, it’s written up on the TensorFoundry blog, pieces like what an LLM proxy actually does and running LLMs on your own infrastructure.

They’re also deliberately lean. I avoid bloated, dependency-heavy software in favour of things that are performant, extensible and maintainable. This is why I don’t write Java - oops, that’s a bit JAR’ing.

I’ve been writing software since 1999, starting with HotHTML, a tool for web developers (C/C++/VB6 for the early versions, C# / .NET WinForms in v4). HotHTML 4 was meant to ship in 2008 and never did. I might still dust off v4 one day. Also in 1999 I joined the Windows NT testing team and ended up deep in Windows Internals, debugging and reverse engineering. That stuck around for years. After Windows 8 the joy went out of it for me, so I moved to Linux full-time (I was already there, just went full freaking Linux).

These days my main focus is AI inference with Rust, Go and Python. Zig is creeping into a few low-level pieces of consulting work, and there’s still .NET around for SixPivot projects.

A note on frontend

I don’t do frontend. I really don’t. (No, I don’t react well to React.) My strength is everything below the browser. But I do love Svelte + SvelteKit and Bun.

Most of my frontend work for products uses AlpineJS or Svelte.

Beyond code

Garden

I grow most of my own greens and post the homegrown adventures on Twitter. Currently self-sufficient on garlic and ginger. The avocados are ridiculously good - oh and corn, sweet corn homegrown is insane!

Dogs

All my machines are named after my pups. 🖥️🐾

Cars

EV enthusiast. I’ve owned every Tesla except the Roadster. After the Model S, nothing really excited me until the Zeekr 7X, which I just took delivery of in October 2025. Currently setting up the Zeekr Insiders group in Australia for fellow owners, and I’m a Zeekr Ambassador for 2026.

About that avatar

The cropped Neo image I use everywhere has been my avatar since 1999. He hasn’t aged a day - unlike me.

This blog

This is where I write up the things I’m building, the rabbit holes I fall into and the bits I want to remember. Mostly systems, embedded, IoT and modern languages (Rust, Go, TinyGo, increasingly Zig). The previous blog ran for about twenty years under the name NERD Thoughts: Not Even Remotely Dorky Thoughts, stolen with thanks from Lisa Simpson. No AI-written articles, just me.

Expect plenty of pop culture references and a love of a good pun. It’s always a pun time, somewhere.

But I have been slack, after dropping Twitter I’ve been quiet.

The site is powered by Hugo with my fork of the hyde-hyde theme.

Need a hand?

AI infrastructure, air-gapped or sovereign inference, fine-tuning, edge AI deployments, that’s what TensorFoundry does. Get in touch.

Connect