Ryan McConville

I'm VP of AI at MITO.ai and an Honorary Professor of AI at the University of Bristol, UK.

I work across AI research and engineering, from the ideas behind a model to the systems that run it at scale.

Ryan outdoors with his dog

Background

At MITO.ai, I lead research and engineering on multimodal generative AI and agentic systems.

Previously, I was CTO and Principal AI Scientist at kluster.ai (acquired by MITO.ai). I led work on distributed inference, fine-tuning, and post-training with reinforcement learning, from system design and evaluation to production deployment.

At Weights & Biases, I worked with ML teams on model evaluation, experiment tracking, and model management.

I completed my PhD in AI in 2017. In 2019, I became a Lecturer (Assistant Professor in US terms) at the University of Bristol. My academic work covers the areas below. I was the main supervisor for seven PhD students, all of whom have completed their PhDs.

Areas I've worked on

  • Graph learningfrom ~2015

    Graph neural networks, community detection, and changing networks.

  • Computer vision & multimodal learningfrom ~2017

    Learning from images and combining different types of data.

  • Reinforcement learningfrom ~2018

    Offline learning and learning from human feedback.

  • Efficient & distributed MLfrom ~2018

    Federated learning, model compression, and learning on edge devices.

  • Representation learningfrom ~2019

    Self-supervised learning, deep clustering, and manifold learning.

Misc writing

(some random) Things I find interesting that don't really fit anywhere else.

  1. Logi Options+ not detecting MX Master 3S for Mac
  2. Milo Contributes to Science
  3. Erdos Number
The misc writing archive

Other things