A few of the projects I've worked on!
Embedded Files
High-Resolution Urban Pollutant Dispersion Modeling (On-Going)
A project born out of mutual interests with the SU2 Foundation and Prof. Nijso Beishuizen of TU/e. We are currently building an urban flow visualization API that couples QGIS with SU2 to simulate pollutant dispersion in Amsterdam using GPU-accelerated atmospheric downscaling. Prof. Marco Giometto of Columbia University was instrumental in identifying the practical challenges in implementation.
Data-Driven Stochastic Generative Model for Subgrid-Scale Closure of Large Eddy Simulations (2025)
Started as a passion project under Prof. Maniarasu Ravi and Prof. Aravindan Vivekanandan at Thiagarajar College of Engineering (TCE) and ended up being my final project at TCE :) . In an MDO context, such generative subgrid models could act as lightweight surrogates that inject missing physics from high-fidelity simulations, making the design loop faster and more reliable under multi-scale uncertainty. The project was among the 10 best B.E. thesis projects (mechanical engineering) in the year 2025.
Accelerating Multi-Scale Fluid-Structure Interaction Using Non-Body-Fitted Cartesian Grids (2024)
An internship project done under Prof. Manikandan Mathur and Prof. Bharath Govindarajan at IIT Madras. Optimized a multi-scale FSI simulation for riverbank/coastal erosion using non-body-fitted Cartesian grids to reduce meshing complexity and computational cost associated with updating fluid-solid interfaces at every time step for complex, deforming geometries.
Adaptive Structural Health Monitoring Using RL-Guided Imaging Optimization (2023)
An internship project done under Dr. Vishwamoorthy & Dr. Nadeem Masood at the Advanced Composites Division, National Aerospace Laboratories (CSIR-NAL). Developed an RL + CV-assisted NDT system that autonomously optimizes camera parameters (shutter speed, voltage gain) to capture maximal-quality images of rapidly propagating cracks, eliminating human error and environmental variability.
Early-Stage Doppler and Missile Tracking Radar Optimization (2024)
An internship project done under Mr. Karthik Tekumalla at Karthik Engineering Pvt Ltd. I developed a parametric MDO module that translated customer requirements into design variables and used preference elicitation to assign weights across competing objectives. Using the exponential weighted-sum formulation from Stanford’s AA222 course, I generated candidate geometries, ran SU2 aero-structural evaluations, and traced the local Pareto set to filter feasible designs. Automating this stage reduced iteration cycles from days to hours and convinced me to establish MDO as my central design methodology Built an automated computational stack for torque and drag optimization for early design stage radars placed to be placed in hill stations. Used Python + GPT-4 API to generate reports and cut early design iteration time by 27%. Watch the first radar built with this module here.
Bio-inspired Flexi-wing for Stall Mitigation in Future Configurations (2023)
A course project for the Engineering Design Course by Prof. Karthikeyan. Characterized the aerodynamic response of the flexi-wing and compared its stall onset and recovery behavior against a rigid baseline configuration. As this was a 5-week project, we did not get to investigate the problem as much as we wanted. Obtained the highest grade in the department (A++ grade).
Restrictor, Intake and Thermal Management System for 390cc Formula Race Car (2023-25)
A project done under Prof. Maniarasu Ravi & Prof. Nijso Beishuizen as an honorary additional academic credit project. Led the Aerodynamics and Systems Integration team. Redesigned the inlet baffle system to shift the stagnation line, introduced micro-vortex generators to re-energize the near-wall flow, and reshaped the curvature to strengthen passive suction through existing low-pressure pockets. These refinements increased cooling efficiency by 23% and reduced system weight by 8 kg.
Aeroelastic Deformation Tracking for Twist Scaling (2022)
A side project of mine where I investigated how small UAV wings exhibit unintentional passive geometric adaptation: natural twisting and camber changes under aerodynamic loads that stabilize flight. To quantify this, I instrumented a 1.4 m wing with strain gauges, IMUs, and airspeed sensors to measure real-time twist, bending, and load response across the flight envelope. These deformation signatures were compared to bird-like passive wing behaviors, revealing how compliant structures delay stall and smooth gusts. The collected data were used to define twist tolerances and derive scaling parameters for wings of different spans, enabling intentional morphing-wing design based on measured aeroelastic response rather than assumptions.
Research Publications and Conferences:
Affordable Collision-Radiative Model with Petrov-Galerkin Model Reduction for Thermochemical Non-Equilibrium Gas: An Expense Comparison and Showcase for Earth and Mars Atmospheres (2025)
Presented at the 67th Annual Meeting of the APS Division of Plasma Physics in Long Beach, California. (Link)
Author: Sai Siddharth (Guided by Prof. Wesley Harris, Charles Stark Draper Professor of Aeronautics and Astronautics, Massachusetts Institute of Technology and Prof. Aravindan Vivekanandan of Thiagarajar College of Engineering)
Description: Thermochemical models and the detailed descriptions of the physics of nonequilibrium phenomena are computationally extremely expensive and are hence out of reach for early plasma researchers around the world. Throughout the years, researchers have modelled empirical formulations to meet the criteria of large-scale, multidimensional simulations. However, the resulting formulations have inherent assumptions that are often inaccurate and fail to capture the original physics. In addition to this, the need for rigorous research and maintenance of such deep formulations requires profound user supervision and a large number of parameter tunings, which broadens the gap for rising researchers to get firsthand complex phenomena-capturing frameworks. In this poster, the author addresses the state-specific thermochemical collision models (collision-radiative) through an efficient model reduction pipeline based on Petrov-Galerkin Projection of the nonlinear kinetic equations onto a low-dimensional subspace. This work is justified by the observation of kinetic systems, which tend to exhibit low-rank dynamics that rapidly move towards a low-dimensional subspace. The reduction is done for two complete atmospheres: Earth and Mars. The calculations of O₂ and N₂ rovibration are presented to show the efficiency of such projections into manageable state variables and the capturing of the complex phenomena in both macroscopic and microscopic quantities. The code can be implemented in a 16 GB RAM system.
Adversarial Trajectory Containment Markov Loop for Hypersonic Non-equilibrium flows (2025)
(Ongoing study) MDPI Aerospace Journal
Authors: Sai Siddharth, Manickam Mahendran, Aravindan V, Maran P
Description: Conceptual design for hypersonic vehicles presents significant challenges due to the coupling between fluid dynamics and structural mechanics, known as Fluid-Structure Interaction (FSI). While projection-based Reduced-Order Models (ROMs) can reproduce the full flow field with approximately 1% macroscopic error, we observed that small deviations from thermochemical equilibrium (TNEQ) can induce order-of-magnitude discrepancies in the predicted aeroelastic loads. Crucially, these discrepancies have the potential to flip conceptual design decisions. This work addresses the need for the design pipeline to anticipate these nonlinear sensitivities inherent in non-equilibrium flow. We propose an adversarial-scenario layer integrated on top of the ROM chain. This layer is designed to identify regimes of thermochemical drift and actively perturb the surrogate model in those directions to stress-test load predictions. Such a module, framed as a trajectory containment loop, can ensure the robustness of the design across a variety of maneuver-prone complexities. Implementing this framework would effectively extend Multidisciplinary Design Optimization (MDO) codes, such as SUAVE, into high-uncertainty design spaces that traditionally require specialized, high-fidelity solvers.
Aerodynamic Shape Optimization - Discrete Adjoint and NURBS for Mesh Deformation (2024)
Presented at the 4th SU2 Conference in Varenna, Italy. (Link)
Author: Sai Siddharth
Precise Pharmacokinetic Optimization of Diabetes Therapeutics (2024-25)
Presented at the 2nd International Conference on Navigation Systems (ICAAN-2025) (Under review at the AAPS PharmaSciTech Journal) (Link)
Authors: Sai Siddharth, Mohan Kumar R (Doctor of General Medicine) and Prof. Maniarasu Ravi
Description: Current oral tablet dissolution tests, such as USP paddle and basket apparatus, rely on global hydrodynamic stirring that produces non-uniform local shear and transient stagnant zones, thereby limiting dissolution efficiency and reproducibility. Therefore, a critical research gap exists in achieving localized, continuous replenishment of the boundary layer at the tablet-fluid interface. In this study, a well-established phenomenon observed in rotating bodies, inspired by the Magnus effect, where the rotation of a tablet generates a strong, localized near-wall shear flow, is introduced for solid dosages. A quasi-2D laminar incompressible flow around a rotating tablet was simulated in OPENFOAM’s transient pisoFoam solver with the snappyhexmesh algorithm for near-wall meshing. The generated asymmetrical flow fields, deflected wakes, and low-pressure regions aid in the rapid surface boundary layer renewal by facilitating enhanced surface convective flux. Existing research references are benchmarked to compare the potential of the hydrodynamic strategy being proposed in this study. Simulations with gastrointestinal/microfluidic conditions reveal dynamic boundary layer disruptions and wake deflection effects, which are not achievable in traditional systems, underscoring the promising route for overcoming hydrodynamic lag associated with conventional apparatus and effectively closing the gap between achieved and target dissolution profiles. This study prioritizes local fluid-structure interactions over bulk stirring to serve as a foundation for next-gen oral dosage forms developed with superior dissolution characteristics and bioavailability.
Splash Dynamics of a Free-Falling Water Droplet (2025)
Accepted to Showcase at the APS Gallery of Fluid Motion (Link)
Authors: Sai Siddharth, Sasidharan Prakasan (University of Houston)
Mini Projects:
RC planes as a hobby!
My very own calibrated wind tunnel!
Fun fact: I'm still calibrating it
600cc engine air intake system
Hexacopter for Off-shore in-situ measurements!
Page updated
Google Sites
Report abuse