My core work is in combustion, propulsion, and multiphysics system simulation. I'm interested in problems where fluid flow, thermodynamics, heat transfer, chemical kinetics, and system behavior interact in complex ways — and where engineering judgment matters as much as mathematical sophistication.
I work on modeling and simulation problems that sit between physical depth and practical usability. The questions that interest me are often of this kind:
How do we represent complex reacting systems with enough physical fidelity to be trustworthy, while keeping the models computationally efficient, interpretable, and useful inside real engineering workflows?
That question leads naturally into areas such as combustion and propulsion system modeling, reacting-flow simulation, chemical kinetics and reduced-order modeling, 0D/1D physics-based models, system-level simulation, and AI as an accelerator for scientific and engineering work.
I'm especially drawn to the translation layer between high-fidelity science and usable engineering models — starting from first principles, identifying dominant mechanisms, and building models robust enough to support design, analysis, and decision-making.
Research
My research follows a thread that started with building things — race cars, propulsion rigs — then went deeper into the underlying science, and is now focused on making that science fast and useful at the system level.
Phase 1 — Formula Student, Baja SAE & Rocket Propulsion (2010–2015)
At IIT Bombay I studied mechanical engineering and spent most of my time building things. As part of the student motorsport teams, I worked on three electric race cars for the Formula Student competition and one all-terrain vehicle for Baja SAE India — projects that covered electric drivetrains, suspension, chassis, and the particular discipline of building complex mechanical systems under time and resource constraints.
For my undergraduate thesis, I worked on developing a lab-scale rocket propulsion research facility at IIT Bombay, funded by ISRO and DRDO. The facility was designed to support experimental combustion and propulsion research — test stands, instrumentation, and the infrastructure needed to safely run small-scale rocket motors. This was my first real exposure to the engineering of propulsion and reacting flows, and it pointed the direction for what followed.
Phase 2 — Energetic Materials: From Molecule to Flame (2015–2019)
My PhD at Penn State under Prof. Stefan Thynell focused on one of the chemically richest combustion problems: the thermal decomposition and combustion of RDX and HMX — the cyclic nitramine compounds that form the energetic core of most modern military explosives and solid rocket propellants.
The central challenge was that the decomposition mechanisms of these materials were not understood at a first-principles level. Prior work relied on fitting global models to experimental data, but without knowing which molecular pathways actually dominate, predictive modeling was limited. My approach was to go upstream: use quantum mechanics calculations — density functional theory and high-accuracy composite methods (G4(MP2), M06-2X) — to map the initial reaction pathways from scratch. Which bonds break first? Which intermediates form? What are the rate-controlling steps?
This produced something new: a comprehensive liquid-phase reaction mechanism derived from first principles, not empirical fitting. The mechanism was then validated against thermolysis experiments — thermogravimetric analysis (TGA) and confined rapid thermolysis (CRT) — with species evolution profiles in good agreement across a range of conditions. Along the way, a novel approach was developed for calculating transport properties directly from intermolecular potentials, eliminating the need for arbitrary bath gas assumptions.
The work culminated in a full multiphase combustion model for HMX monopropellant — liquid phase, gas phase, transport — that correctly predicts burn rates and melt layer thicknesses across pressures. One of the key findings was the role of autocatalytic HONO addition via the cage effect as the dominant pathway for HMX decomposition, a mechanism not previously identified for this material.
Publications:
- Quantum mechanics investigation of initial reaction pathways and early ring-opening reactions in thermal decomposition of liquid-phase RDX — Combustion and Flame, 178, 2017
- Identification of initial decomposition reactions in liquid-phase HMX using quantum mechanics calculations — Combustion and Flame, 188, 2018
- A Comprehensive Mechanism for Liquid-phase Decomposition of HMX: Thermolysis Experiments and Detailed Kinetic Modeling — Combustion and Flame, 212, 2020
- Intermolecular potential parameters for transport property modeling of energetic organic molecules — Combustion and Flame, 200, 2019
- Modeling of HMX Monopropellant Combustion with Detailed Condensed-Phase Kinetics — AIAA Propulsion and Energy Forum, 2019
Phase 3 — Reactive Flow Modeling & Fast Chemical Kinetics (2019–2023)
On joining the Reactive Flow Team at Gamma Technologies, the focus shifted from building new kinetic mechanisms to making existing ones fast and practically useful inside engineering simulation tools.
Detailed kinetic mechanisms are now available for many fuels, but they are computationally intractable in 3D engine CFD: a mechanism with 679 species and thousands of reactions cannot be solved at every cell and every timestep. This phase addressed that bottleneck through two complementary approaches.
The first was a tabulated chemistry method: thermochemical states of the mixture are pre-computed across the relevant operating space using homogeneous reactor simulations, stored in a lookup table, and interpolated at runtime. A novel progress variable formulation and an automatic scheme for identifying important intermediate species made the method robust across a wide range of engine conditions.
750× faster than direct detailed kinetics integration — knock onset predicted to within 0.6 crank angle degrees, cylinder pressure, temperature, CO and NOₓ matched against full detailed chemistry, with computational cost independent of mechanism size.
The second was a machine learning tabulation (MLT) method: deep neural networks trained on pre-computed thermochemical states, clustered by a combustion progress variable. DNNs eliminate the need for table storage entirely, reducing memory requirements by three orders of magnitude while achieving ~300× speedup for a 621-species mechanism.
~300× speedup for a 621-species mechanism with 3 orders of magnitude less memory than a lookup table — making detailed kinetics practical for fast knock detection and engine calibration studies.
Alongside the tabulation work, I also developed and contributed to laminar flame speed solvers, equilibrium chemistry solvers, and non-spherical flame modeling within the GT-SUITE simulation framework — expanding the physical scope of the reacting-flow capabilities in the software.
Publications:
- Tabulated chemistry method for fast calculations of detailed chemical kinetics in system level simulations of internal combustion engines — International Journal of Engine Research, 2022
- Machine Learning Tabulation Scheme for Fast Chemical Kinetics Computation — SAE International Journal of Engines, 2024
Phase 4 — Chemical Reactor Network Modeling for Gas Turbines and Industrial Systems (2023–present)
My current research at Gamma Technologies focuses on Chemical Reactor Network (CRN) modeling — an approach that represents a complex combustion system as a network of idealized, well-stirred and plug-flow reactors, each with specified flow connections and boundary conditions. The appeal of CRNs is that they can incorporate detailed chemical kinetics (the full mechanism, not a surrogate) at a computational cost orders of magnitude lower than 3D CFD, because the spatial problem is collapsed into a network topology.
The applications I am focused on are gas turbines, industrial boilers, furnaces, and process reactors — systems where emissions prediction (NOₓ, CO, unburned hydrocarbons) is critical, where 3D CFD with detailed kinetics is impractical at the design and optimization stage, and where a physically grounded reduced model can enable rapid parametric studies.
CRN modeling sits at a distinctive point in the fidelity spectrum: it is not as cheap as global correlations, but it retains the chemical accuracy needed to predict pollutant formation and fuel flexibility effects. Getting the network topology right — how to partition the combustor volume into reactor zones, how to set inter-reactor flow rates — is the key modeling challenge, and is where much of the interesting work lives.
Full publication list
All publications, citations, and co-author information are on my Google Scholar profile.