Shreshth Malik

Shreshth Malik

Machine Learning DPhil Student @ University of Oxford

OATML

AIMS CDT

About Me

I am a DPhil candidate in Machine Learning at the University of Oxford, supervised by Yarin Gal and Stephen Roberts. I am a member of the OATML group and the AIMS CDT. I recently was a researcher on the NASA FDL-X program.

In general, I am interested in the problems that limit ML systems to be taken out of the lab and used with confidence in the real world, motivated by applications in the sciences. This includes topics such as data-efficient learning, differentiable simulations, and human-AI collaboration. I also think about short/long term AI risks, and collaborate with the causal incentives working group.

Previously, I studied Physics at St John’s College, Cambridge, working on a machine learning approach for predicting the outcomes of material synthesis procedures in the Lee group. I then joined a ML Master’s at UCL, working with Humanloop to combat some of the issues with using active learning in practice. Outside of research, I’ve worked in venture capital, management consulting, and with a range of tech startups.

I am very happy to chat to anyone considering ML careers/PhDs - let me know if I can be of any help. Check my Notion page for some of the resources and advice I have found helpful.

Recent Publications

Quickly discover relevant content by filtering publications.
(2023). High-Cadence Thermospheric Density Estimation enabled by Machine Learning on Solar Imagery. Machine Learning and the Physical Sciences Workshop, NeurIPS 2023.

Preprint Cite

(2023). BatchGFN: Generative Flow Networks for Batch Active Learning. Structured Probabilistic Inference & Generative Modeling Workshop, ICML 2023.

Preprint Cite Code

(2022). Discovering Long-period Exoplanets using Deep Learning with Citizen Science Labels. Machine Learning and the Physical Sciences Workshop, NeurIPS 2022.

Preprint Cite Code

(2022). Multi-Modal Fusion by Meta-Initialization. arXiv preprint arXiv:2210.04843.

Preprint Cite Code DOI

(2021). Predicting the Outcomes of Material Syntheses with Deep Learning. Chemistry of Materials.

Preprint Cite Code DOI