Chris is a machine learning engineer with a focus on computer vision, language modeling, and generative models, and is interested in the intersection of art and machine learning, with a focus on multiplicity, relations to others, queerness, and movement.
Recent artistic collaborations include work with artist Zach Blas on Cultus, an installation commissioned by Arebyte Gallery and Secession, and developing a few-shot gas canister detector for Forensic Architecture.
Chris executed challenging tasks throughout entire pipelines, from dealing with small amounts of data (as little as 10 training data points) all the way to deploying ML models that serve thousands of requests per second.
Chris completed a Master's in Machine Learning at the University of Cambridge in August 2017.
research interests video language models — few shot learning — generative models — Bayesian modeling — computer vision: video understanding, video generation, object and movement detection
other interests
contemporary dance — internal family systems psychotherapy —compassion therapy
CULTUS with Zach Blas — commissioned by Arebyte Gallery, London, UK, and Secession, Vienna, Austria
— text generation Read more︎
Profundior with Zach Blas — commissioned by Berlin Biennale for Contemporary Art, exhibited at Hamburger Bahnhof
— diffusion video generation — text generation Read more︎
2022
576 Tears with Zach Blas — commissioned by UP Projects for “This is Public Space” series
— live GANs video generation — live sentiment camera based detection Read more︎
The Doors with Zach Blas — commissioned by Edith-Russ-Haus für Medienkunst, Oldenburg, de Young Museum, San Fransisco, and Van Abbemuseum, Eidhoven
— GANs video generation — text generation Read more︎
2019
MELTS INTO LOVE with Xin — album cover — neural style transfer
D’Cruz, A.∗, Tegho, C.∗, Greaves, S.∗, & Kermode L. (2022). Detecting Tear Gas Canisters With Limited Training Data. IEEE/CVF
Winter Conference on Applications of Computer Vision (WACV). ∗equal contribution
Tegho, C., Budzianowski, P., & Gašić, M. (2018). Benchmarking Uncertainty Estimates With Deep Reinforcement Learning for Dialogue
Policy Optimisation. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP).
Tegho, C., Budzianowski, P., & Gašić, M. (2017). Uncertainty Estimates for Efficient Neural Network-based Dialogue Policy Optimisation. Accepted at the Bayesian Deep Learning Workshop, 31st Conference on Neural Information Processing Systems (NeurIPS).
machine learning work
2022 - Present Unitary, London, UK — Develop and deploy multimodal machine learning models and pipelines for detecting harmful content in videos, images and text
2017 - 2022 Calipsa, London, UK — Design, implement and evaluate models and software prototypes for object detection and motion detection in videos
Commissioned by Berlin Biennale for Contemporary Art, exhibited at Hamburger Bahnhof
Profundior is an immersive multimedia installation that envisions an AI system as a deity consuming human emotional data. Responding to the quasi-religious veneration of AI in Silicon Valley, it presents an "AI god" in the form of a black box network that ritualistically ingests simulated human tears. Through this speculative work, Profundior interrogates the ethical implications and extractive data practices underlying AI's rapidly advancing emotional intelligence capabilities. In this work, I trained diffusion models on images of tears to generate videos for the installation. My work also involved training language models to generate text for the installation.
Low resolution protype video generated from a diffusion model for Profundior. Final videos only available through the installation.
