Mohith H

Working on multi-scale molecular feature engineering and solvation environment classification for multi-solvent electrolyte systems.

• Engineered a unified molecular feature pipeline using SOAP descriptors across MD trajectories, producing a structured dataset for multi-solvent comparative analysis

• Trained an unsupervised clustering model on high-dimensional SOAP descriptors to classify distinct solvation environments across ACN/DMF trajectories

• Developing MLIP-based methods to predict water activity in multi-solvent electrolyte mixtures using learned solvation geometries

Developing density-derived atomic charge models for long-range electrostatics. Preprint: opt-DDAP (arXiv:2604.10984).

• Built a production-grade Python DDAP pipeline for reciprocal-space Gaussian fitting and stable atomic charge extraction from plane-wave DFT densities, achieving accurate density reconstruction on ionic benchmarks (NaCl vacancy supercells, 7–63 atoms)

• Developed opt-DDAP, a differentiable reformulation of DDAP as a PyTorch computational graph, replacing the numerically fragile Lagrange-multiplier solver with a Moore–Penrose pseudoinverse followed by charge renormalisation — maintaining stability up to condition numbers κ(A) > 10¹⁰

• Enabled gradient-based optimisation of Gaussian basis parameters (σ_start, f, g_c) via automatic differentiation, demonstrating robustness to initial conditions with <2% variation in extracted charges across four distinct starting points

• Validated framework on NaCl vacancy supercells and MoS₂ monolayer, including faithful reconstruction of difference charge densities (Δρ = ρ_defect − ρ_bulk), confirming applicability to defect-induced charge redistribution

Led chemical engineering innovations on net-zero solutions and sustainability projects.

• Completed 2-week trainee program on chemical engineering concepts and sustainability

• Participated in Chem-E-Car Challenge 2025 and ChemE Cube Competition

• Designed complete lab-scale Direct Air Capture (DAC) pilot plant

• Performed detailed HAZOP analysis and proposed safety controls