Statistical mechanics, turbulence, and machine learning.
I study how complex systems become unstable, transition, and organize, from laminar-turbulent flows to optimized neural networks.
My work combines non-equilibrium phase transitions, finite-size scaling, stochastic modeling, and data-driven methods to develop quantitative models for physical phenomena.
Transitional turbulence| Complexity-stability| Scientific ML
About
My PhD work, advised by Nigel Goldenfeld, focuses on the statistical mechanics of turbulence and instability. I use non-equilibrium phase transitions, finite-size scaling, stochastic modeling, and random-matrix ideas to characterize how disordered physical systems transition between stable and unstable states.
Current projects center on tricritical directed percolation in transitional turbulence, the universality class of stratified shear-flow transitions, and May's complexity-stability hypothesis in optimized systems such as neural networks.
Previously: Dual Degree (B.Tech + M.Tech) in Engineering Physics at IIT Bombay — Institute Silver Medal, Best Master's Thesis. Research stints at Aalto University and TIFR Mumbai on quantum condensed matter and topological materials.
News
- Jun 2026 AI/ML Intern at TAU Systems (Carlsbad) — physics-informed ML for laser-plasma electron accelerators.
- Oct 2025 Talk on tricritical DP & transitional turbulence at the JIFT Workshop on Strong Turbulence, UC San Diego.
- Sep 2025 First-author paper published in Phys. Rev. Lett. 135, 104001; covered by UCSD News.
- Mar 2025 Talk at the APS Global Physics Summit, Anaheim CA.
Selected Research
- Tricritical Directed Percolation Controls the Laminar–Turbulent Transition in Pipes with Body Forces
Jayasingh & Goldenfeld. Identifies the tricritical DP universality class governing pipe-flow transition under body forces; reconciles long-standing discrepancies in transition phenomenology.
Finite-size scaling and Binder cumulants resolve whether stratification is a relevant perturbation to DP at the turbulent onset.
Treats trained networks as optimized interaction systems; asks whether SGD-trained dynamics violate random-matrix instability the way evolved ecosystems do.
Toolkit
Contact
gjayasingh@ucsd.edu · gurukalyan1.618@gmail.com
San Diego, CA