0.35 s → 14 s: the TMA pulse that saturates a blanket 300 mm wafer against the pulse a 3D-NAND-like wafer of 100 nm holes at aspect ratio 100 (56× the wall area) needs: a pulse sized on the blanket wafer is 40.7× too short for the patterned one. A showerhead ALD chamber in OpenFOAM in which every wafer face runs its own feature-scale hole model through a C++ boundary condition, two-way coupled to the gas. The resulting supply rule, t_pulse = 1.2 (1 + A_wall/A_wafer) N₀/(U c₀), holds within 25 % over aspect ratios 20 to 100 and a 20× range of TMA feed across 12 chamber runs; the 1D hole model sits within 3 to 11 % of a converged ballistic reference (Argonne's Machball and the repository's own 3D voxel Monte Carlo solver, which agree within 2 %), fits published PillarHall profiles, and the rule puts a measured fluidized-bed ALD run on 940 m² of silica gel at 1.12× the supply limit. 25-run central composite design (JMP response surface, R² 0.993), Gaussian-process surrogates in MATLAB and scikit-learn (LOO R² 0.999), utilization 51 to 97 %, 12 % precursor saving per wafer at the same pulse time. OpenFOAM C++ boundary conditions JMP MATLAB scikit-learn Repository · Live page

Spearman ρ = 0.76 (p = 0.007) on 11 experimentally characterised cathodes. Built an eight-layer funnel over 2,774 Li-ion insertion electrodes from the Materials Project and a three-descriptor Degradation Screening Index that outranks volume strain alone (0.55) and puts 10 of 11 cathodes on the right side of the durable split. Sole author; preprint on ChemRxiv, manuscript under revision at Phys. Chem. Chem. Phys. Python pymatgen scikit-learn SHAP Repository · Live page

26 mV training RMSE with three fitted parameters; the unseen 2C average voltage predicted within 17 mV. Cycled commercial LMO half cells on a Bio-Logic VMP-300 (CV at 0.1 to 0.3 mV/s; C/10 to 2C), fitted a Doyle-Fuller-Newman model on C/5 and C/2 only, validated it blind at 2C and against all six CV peaks (within 80 mV), and named the 2C loss: 74 % ohmic, 17 % charge transfer, 6 % diffusion. 58 automated tests. PyBaMM SciPy pytest Repository · Live page

Wafer non-uniformity from 179.6 % to 3.3 %. Built a 300 mm single-wafer showerhead reactor (2D axisymmetric: flow, heat, species, surface kinetics), diagnosed the starved deposit as mass-transfer limited (Da above 10), fixed it at the inlet (10 to 150 mm gives 6.1 %) and reached 3.3 % at 0.07 m/s with a 14-run central composite design. Gaussian-process surrogate with leave-one-out R² 0.944, 10,000-draw Monte Carlo on MFC and thermocouple tolerances, 3D model with asymmetric exhaust at 5.9 %. COMSOL DOE Gaussian process Monte Carlo Repository · Live page

[OMIM]⁺ binds at −8.75 kcal mol⁻¹ per ion. All-atom NAMD simulations of four imidazolium [BF₄] ionic liquids confined in a (15,15) nanotube showed that chain length sets orientation, binding and mobility at the wall. Co-author, Structural Chemistry (Springer Nature, 2024), doi 10.1007/s11224-024-02323-3. The repository ships the 19,260-atom minimised structure, which the live page renders. NAMD OPLS-AA VMD Repository · Live page

27.99 % simulated efficiency with a Cs₂BiAgI₆ electron-transport layer and nickel contact (V_OC 1.21 V, J_SC 25.96 mA cm⁻², FF 89.3 %), from a six-parameter design of experiments over ten electron-transport layers and nine back-contact metals on an FTO/ZnSe/CdTe heterojunction. Spiral panel architecture with 17.9 % more active area: two granted Indian design registrations, one utility application under review. Best Poster Award, 61st Annual Convention of Chemists. SCAPS-1D drift-diffusion device physics Repository · Live page

A sharp cascade threshold in through-layer conductivity between 1.0 and 1.25 W m⁻¹ K⁻¹. Built a 20-cell 5s4p pack of 4680 cells on COMSOL's validated propagation model (25 Ah, 110 kJ per cell, 80 °C trigger) and ran twelve transient one-hour studies over base cooling from 0 to 2,000 W m⁻² K⁻¹: below the threshold about 13 of 20 cells cascade within the hour, above it propagation stops at the initiating cell; the threshold is a property of the cell rather than the cooling, and internal conduction, not the base contact, is the controlling resistance. Reduced-order model and tested scaling analysis in the repository. COMSOL Battery Design Module reduced-order model pytest Repository · Live page