Sri Hari Krishn Narayanan
University of Chicago, IL·Computer Science
Simulations of the Earth, from ocean currents to flowing ice, depend on parameters whose values are uncertain. Figuring out how sensitive a simulation is to each parameter requires derivatives, the mathematical quantities that describe how an output changes when an input is nudged.
- Fields
Computer Science
Earth & Environmental Science
Mathematics & Statistics
- Topics
automatic differentiation
differentiable programming
Julia
- Grants
- 1 known grant, through Jul 2027
What they work on
- Develop an open-source, general-purpose automatic differentiation framework in Julia that computes exact derivatives of program code for both large-scale Earth system models and scientific machine learning methods.
- Enable universal differentiable programming that unifies simulation and machine learning algorithms, including physics-informed neural networks and neural differential equations, for hybrid Bayesian inversion and Bayesian machine learning.
- Deliver derivative-enabled flagship applications, an open-source ocean general circulation model from the Climate Modelling Alliance and an open-source ice flow model, with gradient and Hessian information available to the community.
- Support parameter and state estimation, data assimilation, Hessian-based and gradient-informed MCMC uncertainty quantification, and quantitative observing system design using the differentiated Earth system models.
More about the research
The same derivatives underlie training neural networks and Bayesian inference, the statistical approach to estimating unknowns with uncertainty. This collaborative project builds software so that Earth system models and machine learning methods can share one derivative engine, letting the two fields be combined rather than used separately.
Why it matters
Climate projections and ice-sheet forecasts hinge on how well models are calibrated to observations and how honestly their uncertainty is reported. Open-source differentiable ocean and ice flow models could make parameter estimation, data assimilation and observing-system design more tractable for a wide community, and partners are already applying the same tools to power grid efficiency, personalized drug dosing and EEG instrument design.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Patrick Heimbach
University of Texas at Austin, TX·Earth & Environmental Science
Set deep-ocean observing requirements by working with Global Ocean Observing System expert panels to refine Essential Ocean Variables, and apply them through regional demonstration projects.
Jonghyun Lee
University of Hawaii at Manoa, HI·Earth & Environmental Science
Build hybrid surrogate models from neural operators, networks that imitate expensive simulations, coupled through an open-source framework so groundwater–ocean exchange is simulated quickly while conserving mass at the land–sea boundary.
Burak Aksoylu
Texas A&M University-San Antonio, TX·Mathematics & Statistics
Use machine learning to improve iterative local-to-nonlocal coupling, so that peridynamics models only the region where fracture occurs and cheaper local models handle the rest.
The school
University of Chicago: 39 labs on PI Pages, $681M a year on research, 457 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.