Robert McHardy

Robert McHardy

Member of Engineering (Applied Research)

poolside

Biography

I am part of the pre-training team at poolside inside the Applied Research organisation, where I work on pre-training large language models for code generation.

Previously, I was a Senior Researcher in Natural Language Processing (NLP) and Automatic Speech Recognition (ASR) at AssemblyAI and an AI Research Engineer at InstaDeep’s and BioNTech’s joint AI innovation lab, where I led a team of four engineers to develop AI models for the treatment of cancer and prevention and therapy of infectious diseases, including SARS-CoV-2.

I have a Master of Science in Machine Learning from University College London, where I was part of the Deciding, Acting, and Reasoning with Knowledge (DARK) lab, being supervised by Mikayel Samvelyan, Patrick Lewis, and Tim Rocktäschel. Before joining UCL, I completed my Bachelor of Science in Natural Language Processing at the University of Stuttgart under the supervision of Roman Klinger.

Interests
  • Artificial Intelligence
  • Natural Language Processing
  • Speech Understanding
  • Graph Neural Networks
Education
  • MSc in Machine Learning, 2021

    University College London

  • BSc in Natural Language Processing, 2019

    University of Stuttgart

Experience

 
 
 
 
 
poolside
Member of Engineering (Applied Research)
Oct 2024 – Present London, United Kingdom
Pre-training large language models for code generation.
 
 
 
 
 
AssemblyAI
Senior Researcher
Feb 2024 – Oct 2024 London, United Kingdom
Training multimodal (text and speech) large language models in JAX.
 
 
 
 
 
AssemblyAI
Researcher II
Sep 2023 – Feb 2024 London, United Kingdom
Led self-supervised pre-training research of SOTA multilingual speech-to-text model on 12.5M hours of audio (~4.5T tokens) for models ranging from 660M to 2B parameters, trained on 512 TPU v5e. Implemented and optimised paired Jax and PyTorch codebases for model training and deployment.
 
 
 
 
 
InstaDeep
Research Engineer II
Oct 2021 – Aug 2023 London, United Kingdom
Led a team of four research engineers, developing, evaluating and deploying state-of-the-art AI models (Transformers and Graph Neural Networks) based on the latest AI research applied to biology.
 
 
 
 
 
Bosch Center for Artificial Intelligence
Machine Learning Research Intern
May 2020 – Aug 2020 Renningen, Germany
Developed a state-of-the-art relation extraction model based on distributional similarity with TensorFlow and transformers. Generated an automatically annotated text corpus with unique named entity identifiers based on the English Wikipedia encyclopedia.
 
 
 
 
 
ComputerBase GmbH
Freelance Editor
Sep 2014 – Present Berlin, Germany
Researching and reporting about current IT topics including artificial intelligence for one of the largest online computer magazines in Germany (>30 million PIs per month).

Publications

(2024). Are We Done with MMLU?. In arXiv preprint.

PDF Cite DOI

(2024). Anatomy of Industrial Scale Multilingual ASR. In arXiv preprint.

PDF Cite DOI

(2023). LightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks. In NeurIPS MLSB 2023.

Cite DOI

(2023). Challenges and Applications of Large Language Models. In arXiv preprint.

PDF Cite DOI

(2022). Peptide-MHC Structure Prediction With Mixed Residue and Atom Graph Neural Network. In NeurIPS MLSB 2022.

Cite DOI

(2019). Adversarial Training for Satire Detection: Controlling for Confounding Variables. In NAACL 2019.

PDF Cite DOI