Shrenik Zinage

Internships

May 2025 – Dec 2025

Part-time Student Researcher/Research Scientist Intern — Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA (USA)

Estimated and calibrated large scale commercially available first principles dynamic models of vapor compression cycles (VCCs) in Julia for robust control and fault detection applications.
Proposed a sequential Bayesian calibration framework for VCC systems that explicitly quantifies various sources of uncertainty in model predictions to improve model generalizability across datasets.
Investigated refrigerant mass leakage in aging VCC systems using the proposed approach.
Patent filing and research paper submissions are in progress.

May 2024 – Aug 2024

Physics‑Informed Machine Learning Intern — RTX Technology Research Center, East Hartford, CT (USA)

Introduced a scalable deep kernel using kolmogorov arnold networks as an effective alternative to deep kernel learning using multi layer perceptrons.
Developed an information field theoretic approach for learning partial differential equations using neural operators.
Conducted research toward pretrained foundational models for physics and engineering.

May 2019 – Jun 2019

Industrial Intern — Mazagon Dock Shipbuilders Limited, Mumbai (India)

Worked in production and workflow processes.
Worked closely with public sector entities with respect to quality control measures followed in MDL.

Research Projects

Causal Digital Twins for Real-Time Decision Making and Fault Detection [Sep 2024 – Present]

Developed a novel scalable approach to causal discovery.
Developed an AI agent for automating end to end causal inference.
Evaluating the real world applicability and effectiveness of our proposed approach to fault detection in a real world engineering scenario.

Skills: Digital Twins, Causal Discovery

Project image: Causal Digital Twin for Real-Time Decision Making and Fault Detection

Probabilistic System Modeling and Adaptability for Engine‑out NOx [Jan 2023 – Present]

Developed a predictive probabilistic model for engine‑out NOx using Gaussian process regression.
Incorporated physical laws in the deep kernel with a causal graph derived via graph convolutional networks.
Quantified the epistemic and aleatory uncertainty in NOx predictions.
Assessed the performance of the model across quantitative and qualitative metrics.
Built a stable linear control oriented model for predicting engine-out NOx using deep Koopman operator.
Proposed a Bayesian calibration framework combining Gaussian processes with approximate Bayesian computation to build transferable probabilistic models for engine-out NOx.

Skills: Graph neural networks, Digital twins, CNNs, Deep Koopman, Sobol analysis, Engine modeling, Gaussian processes, Causality, Bayesian calibration
Libraries: GPyTorch, JAX, Pyro, NumPyro, SALib
Links: Paper 1 | Paper 2

Project image: Probabilistic System Modeling and Adaptability for Engine‑out NOx

Multiscale Models for the Mechanical Response of FCC Alloys under High Strain Rates and Complex Triaxial Loads [Aug 2022 – Present]

Created an innovative multiscale model for predicting metallic alloys response to high strain loads using physics informed neural networks.
Merged large scale atomistic simulations with machine learning for optimal model development.
Acheived improved results, capturing various deformation mechanisms under extreme conditions..

Skills: State‑space modeling, Recurrent neural networks, Hypersonics, Material science, Physics‑informed ML
Libraries: PyTorch

Links: Paper in progress

Project image: Multiscale Models for the Mechanical Response of FCC Alloys

TransformerMPPI: Accelerated Model Predictive Path Integral Control with Transformer‑Initialized Control Sequences [Oct 2024 – Dec 2024]

Integrated transformers with MPPI control to improve computational efficiency.
Used a trained transformer to produce informed mean control sequences, reducing sample requirements and accelerating convergence.
Acheived superior performance in collision avoidance and autonomous racing, outperforming MPPI in cost, speed and efficiency.

Skills: Transformers, MPPI
Libraries: PyTorch
Links: Paper | Github code

Scalable Deep Kernel Learning using Kolmogorov‑Arnold Networks [May 2024 – Aug 2024]

Introduced a scalable deep kernel learning (DKL) using Kolmogorov-Arnold Networks (DKL‑KAN) as an effective alternative to DKL using multilayer perceptrons (DKL-MLP).
Analyzed two variants of DKL-KAN for a fair comparison with DKL-MLP: one with same number of neurons and layers as DKL-MLP, and another with approximately same number of trainable parameters.
Used kernel interpolation for scalable structured Gaussian processes (KISS-GP) for low-dimensional inputs and KISS-GP with product kernels for high-dimensional inputs.
Efficacy of DKL-KAN was evaluated in terms of computational training time and test prediction accuracy across a wide range of applications.

Skills: KAN, Deep kernel learning, Gaussian processes
Libraries: GPyTorch
Links: Paper

Project image: Scalable Deep Kernel Learning using Kolmogorov‑Arnold Networks

Leveraging Gated Recurrent Units foe Iterative Online Precise Control for Geodetic Missions [Jan 2024 – Apr 2024]

Developed an innovative iterative modification to PID controllers for improved attitude control precision in geodetic missions like GRACE-FO.
Leveraged GRU to predict external disturbances from satellite attitude measurements, improving the standard PID control loop.
Demonstrated significant reduction in attitude error through simulation, validating the effectiveness of the proposed GRU-augmented PID approach.

Skills: GRU networks, PID control
Libraries: JAX
Links: Paper | Github code

Project image: Leveraging Gated Recurrent Units for Precise Attitude Control for Geodetic Missions

Liquid Time Constant Networks for Engineered Systems [Jan 2022 – Aug 2022]

Evaluated and demonstrated the superior performance of Liquid Time-Constant (LTC) networks in learning dynamics from noisy data compared to traditional recurrent neural networks.
Used synthetic, corrupted data to test the robustness of these networks under various conditions and parameter settings.
Showcased the effectiveness of LTC networks in modeling standard oscillatory systems under diverse test excitations.

Skills: Liquid time‑constant networks, Neural ODEs
Libraries: PyTorch
Links: Part of MS thesis

Project image: Liquid Time Constant Networks for Engineered Systems

Data Driven Modeling of Turbocharger Turbine using Koopman Operator [Aug 2021 – Jan 2022]

Developed a predictive model for transient and steady-state behavior of a turbocharger using the Koopman operator approach, significantly improving the understanding and control design of the system.
Applied Extended Dynamic Mode Decomposition to approximate the action of the Koopman operator, using experimental data from a Cummins heavy-duty diesel engine, outperforming existing nonlinear autoregressive models with exogenous inputs.
Utilized enhanced sensor data for more accurate modeling, addressing gaps in manufacturer-provided maps, especially in wide operating regions and incorporating heat transfer effects for more comprehensive and realistic modeling.

Skills: Koopman operator, NARX, Turbocharger modeling, MATLAB
Links: Paper

Project image: Data Driven Modeling of Turbocharger Turbine using Koopman Operator

Reinforcement Learning based Control for Active Heave Compensation [Sep 2020 – Aug 2021]

Implemented Deep Deterministic Policy Gradient (DDPG) algorithm to capture the experience of the RL agent during training trails.
The simulation results demonstrated upto 10% better heave compensation performance of RL controller as compared to a tuned Proportional-Derivative Control.
The performance of the proposed method was compared with respect to heave compensation, offset tracking, disturbance rejection, and noise attenuation.

Skills: RL, DDPG/DQN, Python
Links: Paper | Github code

Project image: Reinforcement Learning based Control for Active Heave Compensation

Data Driven Control for Active Heave Compensation [Apr 2020 – Nov 2020]

Performed the model identification of the winch model using long short term memory (LSTM) based recurrent neural network.
The neural network was trained using pairs of input-output data where the input was the opposite of the net heave time history of an KCS ship and output was the control input to winch placed on board the ship.
In addition, the ability of the LSTM network in handling the hard constraints of the swash angle was also analysed.

Skills: LSTM networks, Data driven control, MATLAB
Links: Part of B.Tech thesis

Project image: Data Driven Control for Active Heave Compensation

A Comparative Study of Different Active Heave Compensation Approaches [Aug 2019 – Aug 2020]

Analyzed and compared the performance of various control strategies in keeping a suspended payload regulated from a KCS container ship while the vessel is subjected to changing sea conditions.
The control strategies implemented were Proportional-Derivative (PD) control, Model Predictive control (MPC), Linear Quadratic Integral (LQI) compensator and Sliding Mode control (SMC).
Simulations were performed in MATLAB/SIMULINK environment for three scenarios: no disturbance or measurement noise, with disturbance but no measurement noise and with measurement noise but no disturbance.

Skills: PID control, MPC, Sliding mode control, LQI control, Active heave compensation, MATLAB
Links: Paper | Github code

Project image: A Comparative Study of Different Active Heave Compensation Approaches

Skills

Programming Languages

Python Julia MATLAB/Simulink C++ Java R Fortran

Programming Environments

Git LaTeX ROS Docker Slurm

Software & Libraries

JAX PyTorch TensorFlow Keras GPJax/GPyTorch Flax Equinox Jraph Diffrax Optax Pyro PyMC SALib SciML.jl DeepXDE Neuromancer ModelingToolkit.jl Langchain Langgraph

Computer-Aided Design Software

AutoCAD Fusion 360 Ansys OpenFOAM InkScape

Courses

Graduate Coursework
(Purdue University, USA)

Generative Models (Fall 2025)
Foundations of Computational Imaging (Fall 2025)
Topics in Advanced Scientific Machine Learning (Spring 2024)
Numerical Methods (Fall 2023)
Advanced Mathematics for Engineers and Physicists – 1 (Fall 2023)
Advanced Thermodynamics (Spring 2023)
Advanced Mathematics for Engineers and Physicists – 2 (Spring 2023)
Introduction to Probability Theory (Fall 2022)
Bayesian Data Analysis (Fall 2022)
Introduction to Scientific Machine Learning (Summer 2022)
Linear Algebra (Spring 2022)
Statistical Methods (Spring 2022)
Intermediate Heat Transfer (Fall 2021)
Theory and Design of Control Systems (Fall 2021)

Undergraduate Coursework
(IIT Madras, India)

Advanced Linear Control Systems
Control Engineering
Differential Equation
Guidance and Control of Marine Vehicles
Introduction to Robotics
Machine Learning
Marine Robotics
Modern Control Theory
Nonlinear Programming
Parameter and State Estimation
Probability and Stochastic Process
Series and Matrices
Ship Dynamic Positioning Systems
Ship Hydrodynamics
Ship Motion and Control
Wave Hydrodynamics

Publications

Bayesian Calibration of Engine-out NOx for Engine-to-Engine Transferability

Shrenik Zinage, Ilias Bilionis, Peter Meckl

Accepted at at International Journal of Engine Research

Paper

First page of paper: Bayesian Calibration of Engine-out NOx for Engine-to-Engine Transferability

Sequential Model Calibration of Vapor Compression Cycles using Approximate Bayesian Computation

Shrenik Zinage, Vaibhav Dixit, Ankush Chakrabarty, Hongtao Qiao, Christopher Laughman, Ilias Bilionis, Vedang Deshpande

Accepted at American Control Conference 2026

First page of paper: Sequential Model Calibration of Vapor Compression using Approximate Bayesian Computation

A Causal Graph-Enhanced Gaussian Process Regression for Modeling Engine-out NOx

Shrenik Zinage, Ilias Bilionis, Peter Meckl

International Journal of Engine Research

Paper

First page of paper: A Causal Graph-Enhanced Gaussian Process Regression for Modeling Engine-out NOx

Transformer-Based Model Predictive Path Integral Control

Shrenik Zinage, Vrushabh Zinage, Efstathios Bakolas

Paper
Code

First page of paper: Transformer-Based Model Predictive Path Integral Control

DKL‑KAN: Scalable Deep Kernel Learning using Kolmogorov‑Arnold Networks

Shrenik Zinage, Sudeepta Mondal, Soumalya Sarkar

Leveraging Gated Recurrent Units for Precise Attitude Control for Geodetic Missions

Vrushabh Zinage, Shrenik Zinage, Srinivas Bettadpur, Efstathios Bakolas

AIAA SciTech 2025

Paper
Code

First page of paper: Leveraging Gated Recurrent Units for Precise Attitude Control for Geodetic Missions

Data Driven Modeling of Turbocharger Turbine Using Koopman Operator

Shrenik Zinage, Suyash Jadhav, Yifei Zhou, Ilias Bilionis, Peter Meckl

IFAC PapersOnline (2022)

Paper

First page of paper: Data Driven Modeling of Turbocharger Turbine Using Koopman Operator

Deep Reinforcement Learning Based Controller for Active Heave Compensation

Shrenik Zinage, Abhilash Somayajula

IFAC PapersOnline (2021)

Paper
Code

First page of paper: Deep Reinforcement Learning Based Controller for Active Heave Compensation

A Comparative Study of Different Active Heave Compensation Approaches