IAI Platform
I spent 4 years transforming a hacked-together Airflow pipeline into a highly scalable machine learning platform capable of processing several terabytes of microscopy data daily, enabling the biology team to iterate quickly.
I bulid things
From highly scalable Kubernetes clusters and code sandboxes to CNC programs for robot parts, and many things in between.
I like to solve challenging technical problems. Currently I'm in SF. I spent a couple years vagabonding around the planet before moving here. I enjoy working on Kubernetes clusters, and scaling compute. Over the past decade I have gone from building file servers for small businesses, to writing the firmware in solar cars, helping accelerate drug discovery by building highly scalable compute clusters in the cloud, scaling code sandboxes for AI agents, and now running production infrastructure at Contour.
When I'm not coding, you can find me stargazing, riding my motorcycle, portaging in the wilderness, or scuba diving.
Built the company's production AWS/Kubernetes platform from scratch so the rest of the team could ship without thinking about the cluster.
Scaled the code sandbox platform that lets the Julius agent actually run code.
Built a bunch of different things for clients while vagabonding around the planet.
Scaled the ML pipeline from Phenomic to handle tens of terabytes of data, and to turn around results in a much quicker amount of time.
Built a highly scalable machine learning platform to accelerate drug discovery.
Built and raced a solar car across the Australian Outback.
I spent 4 years transforming a hacked-together Airflow pipeline into a highly scalable machine learning platform capable of processing several terabytes of microscopy data daily, enabling the biology team to iterate quickly.
Working primarily on the firmware team I built the embedded firmware for the solar car — driver inputs to the motor, plus all of our telemetry. I also ran the CFD compute cluster, wrote the CNC programs for the chassis mould, and drove the car.
Got mediated GPU virtualization working on an Intel integrated GPU — GVT-g on KVM/QEMU with VFIO/mdev — so multiple guests could share one iGPU instead of taking exclusive passthrough. A long Gentoo kernel-config rabbit hole that actually worked.
Please reach out if you have something interesting to talk about.