Matt Henderson
VP Of Research at PolyAI
City of Edinburgh, Scotland, United Kingdom
Summary
👤
Senior🎓
Top SchoolMatt Henderson is a VP of Research and conversational AI specialist with a decade of experience building production-grade language understanding systems and leading teams at the intersection of research and product. He has driven voice and dialog innovations from academic PhD work at Cambridge through roles at Google Research, Apple Siri, and multiple PolyAI stints, and most recently leads generative-AI efforts for customer support agents. Comfortable both in C++ tokenizer kernels and large-scale model training, he contributed robustness improvements to tensorflow/text that strengthened real-world tokenization. A founding researcher at frontier model shop Reka AI, he has hands-on experience across post-training, RLHF, retrieval, evaluation and multimodal LLMs. Based in Edinburgh with a background in mathematics and speech processing, he blends rigorous theory with pragmatic engineering to ship ML systems that work in noisy, conversational settings. Colleagues know him for turning subtle research insights into reliable production features rather than just papers.
10 years of coding experience
7 years of employment as a software developer
MSc, Speech and Language Processing, MSc, Speech and Language Processing at The University of EdinburghPhD, Spoken Dialogue Systems, PhD, Spoken Dialogue Systems at University of Cambridge
Github Skills (9)
tokenize10
tokenizer10
nlp10
tensorflow10
python10
natural-language-processing10
cprogramming-language9
machine-learning9
c-language9
Programming languages (8)
Github contributions (5)
tensorflow/text
Aug 2019 - Aug 2019
Making text a first-class citizen in TensorFlow.
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:28 commits, 2 PRs, 22 comments in 21 days
Contributions summary:Matt primarily focused on enhancing the `tensorflow/text` repository's text processing capabilities by adding support for splitting unknown characters within the wordpiece tokenizer. They modified the core kernels of the tokenizer and updated the related test suite to validate the new functionality. The changes include modifications to the C++ code for the tokenization logic, and adjustments to the Python test files to cover the new features. These additions improved the tokenizer's robustness and accuracy in handling diverse text inputs.
tensorflow
matthen/matthen
Mar 2018 - Jul 2023
personal website
Contributions:2 PRs, 68 pushes, 6 branches in 5 years 4 months
reactnextjstypescript
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.