Skip to content

Experience

LLM inference on Kubernetes. vLLM, NVIDIA Dynamo, GPUs that should be working harder than they are.
CUDA enjoyer in the evenings. Nix, Neovim and a terminal person always. Sometimes I give talks about standing this stuff up on EKS.

  • Python
  • Dynamo/vLLM
  • LLM Inference
  • Kubernetes
  • Platform Engineering
  • AWS
  • Nix
  • Linux

Professional Experience

  1. Likezero

    Sept 2024 – Present

    Senior AI Engineer

    • Own and architect Likezero's AWS Kubernetes LLM inference platform, built from the ground up. vLLM and NVIDIA Dynamo, prefill/decode disaggregation, KV-cache-aware routing, speculative decoding and GPU node autoscaling with Karpenter and KEDA.
    • Cut LLM cold-starts by 80%, raised KV cache hit rates and 2-4x increased tokens decode throughput. Served Llama 3.3, Qwen 3.6/3.8, Deepseek v4 and embedding models on RTX 6000 Blackwells and H100s and H200s.
    • Platform is declarative GitOps (ArgoCD, Terraform) with cluster networking (Cilium), TLS (cert-manager), observability (Loki, Prometheus and Grafana) and company-wide services.
    • Built a high-concurrency asynchronous Python server bridging RabbitMQ and HTTP. This underpins Likezero's decoupled autoscaling infrastructure by enabling RabbitMQ/KEDA-driven scaling. This has delivered hundreds of millions of LLM and internal service requests.

  2. Applied Data Science Partners

    Aug 2023 – Sept 2024

    Senior AI Engineer

    • Three projects delivered with LLMs in defence and education. Used many closed and open-source models, such as GPT, Gemma, and Llama, in addition to RAG, chain-of-thoughts, and Retrieval Augmented Thoughts.
    • Led a gen AI and software engineering project that uses decoder transformers for novel use cases in cyber defence.
    • Author of an AI-centred newsletter. With over 850 subscribers, the newsletter has furthered the company's reputation and consulting opportunities.

  3. Liveminds and Maeving

    Jul 2023 – Sept 2024

    AI & Data Engineer Consultant

    • At Liveminds, I developed and productionised a text summarisation product with a multi-step RAG process using GPT and Claude. This kickstarted Liveminds' AI product offering and attracted investor attention.
    • At Maeving, I created a full ETL and analytics pipeline with Azure, Postgres, and PowerBI. These systems led to deepened analytics across the company and self-serve from a trusted single data source.

  4. Laing O'Rourke

    Jun 2021 – Jul 2023

    Data Scientist

    • Tasked with providing economic strategy for national infrastructure projects, I delivered a complete end-to-end data product for multi-billion pound financial data using Azure ML.
    • Before the advent of GPT-3.5, I created a system to understand construction-specific language by fine-tuning a Sentence BERT model. This system created future opportunities for NLP modelling in the broader construction industry.
    • Fostered company data culture by reporting on exciting news in data, presenting and teaching Python/data analytics skills.

  5. Faceit

    Sept 2020 – June 2021

    Junior Data Scientist

    • By analysing the matches of over 2 million players, I created predictive models to determine player's skill to solve the platform's 'smurf' problem.
    • I improved a model that detects toxicity in millions of monthly chat messages. My work improved the platform experience for millions of users.
    • I provided insight into all business areas by creating dashboards and datamarts within Looker and BigQuery.

  6. Various Companies

    Oct 2018 – Aug 2020

    Freelance Data Scientist

    • Created and deployed a predictive model that led to a 50% decrease in costs associated with advertising and employee time.
    • Clarified trends and statistical phenomena with various data by providing business intelligence across multiple companies.
    • Designed and productionised a system that could automatically scrape, clean, and add data to a database from the NHS jobs website.

Education

Keele University

Sept 2016 – July 2019

BSc (Hons) Physics and Mathematics

Talks

Building Inference Architecture with EKS

25th September 2025

AWS

Accelerating GPU workloads on AWS: from GenAI to HPC

Talk at AWS on how to deploy LLMs for contract intelligence application on AWS, highlighting real-world architecture decisions, the journey to building a scalable, cost-efficient inference platform with Amazon EKS and GPU instances on AWS.