Oli Wilkins
Senior Inference Engineer
I build LLM inference platforms on Kubernetes.
vLLM, NVIDIA Dynamo, and GPUs that should be working harder than they are.
Experience
Likezero
Sept 2024 – PresentSenior AI Engineer
Own and architect Likezero's AWS Kubernetes LLM inference platform, built from the ground up. Cut LLM cold-starts by 80%, raised KV cache hit rates and 2-4x increased tokens decode throughput.
Applied Data Science Partners
Aug 2023 – Sept 2024Senior AI Engineer
Three projects delivered with LLMs in defence and education. Led a gen AI and software engineering project that uses decoder transformers for novel use cases in cyber defence.
Liveminds and Maeving
Jul 2023 – Sept 2024AI & Data Engineer Consultant
At Liveminds, I developed and productionised a text summarisation product with a multi-step RAG process using GPT and Claude. At Maeving, I created a full ETL and analytics pipeline with Azure, Postgres, and PowerBI.
Projects
h8s
181 GitHub starsHomernetes is a Talos OS based K8s cluster for my homelab.
argo-workflowsargocdcert-managerciliumgrafanahomelabkubernetesprometheustalos-linux.dotfiles
42 GitHub starsMy personal Nix dotfiles for Linux and MacOS.
configdotfilesneovimnixshellzshinferlace
0 GitHub starsOne model inference engine written by hand for RTX 4080 Lovelace GPU architecture.
cudagpuinferencellmpythonpytorch
Services
A few things I host on my Kubernetes homelab.