Debangshu Mukherjee · Materials scientist
I build physical models and computational tools for understanding materials measurements. My experimental experience spans electron microscopy, optical microscopy, and RHEED measurements. I combine that hands-on work with inverse problems, GPU computing, and interfaces that let AI agents run simulation and reconstruction workflows.
Differentiable physical models
I built diffpes to calculate photoemission spectra from electronic structure, including polarization, broadening, and instrument effects. Its numerical kernels support derivatives for fitting continuous parameters.
Shown: simulated graphene bands and ARPES intensity.
Explore diffpes ↗Quantitative materials characterization
In my first-author ACS Catalysis study, we used 4D-STEM to measure lattice strain in rhodium–platinum core–shell nanocubes with sub-picometer precision.
My contribution: First-author research on quantitative strain measurement using electron diffraction.
Read the published study ↗Scientific computing at scale
2.2 minutes
Published reconstruction time for a dataset with 16,632 probe positions on 4,158 GPUs in our SC22 study.
My contribution: Coauthor of the distributed ptychography study. These measurements describe the team’s benchmark.
Explore the team’s result ↗I build tools for wave optics, electron scattering, surface diffraction, and photoemission. The same physical models support simulation and parameter fitting.
If you’re working on AI for materials, scientific software, or experimental methods, you can reach me at contact@debangshu-mukherjee.com. You can also find more about my background in my resume and CV.