Undergraduate Research Intern
Indian Institute of Technology, Patna
Department of Computer Science and Engineering, IITP • Remote
- Developing an LLM and reinforcement learning-based system to summarize patient-reported adverse drug events in cancer treatment.
- Designing NLP methods to process patient-reported data and generate concise, interpretable summaries for clinical decision-making.
- Investigating scalable pipelines for continuous model improvement and integration into real-world healthcare settings.
- Skills: Natural Language Processing, Large Language Models, Reinforcement Learning, HuggingFace, PyTorch
Undergraduate Project Intern
Autonomous Systems Laboratory, IIT Madras
Department of Engineering Design, IITM • Chennai, Tamil Nadu, India
- Utilized Altair MotionSolve to develop and validate dynamic simulations for a company-sponsored project, analyzing stability under varying loads and terrains.
- Simplified and optimized a CAD model with 10+ components, reducing computational load while maintaining accurate mass distribution.
- Simulated cornering and slope tests of a 1.6-ton self-propelled boom sprayer, identifying critical speed (≈41 km/h) and slope angle (≈30°–33°) for stability.
- Delivered design insights on center of gravity placement and stability margins, informing safer and more reliable off-road agricultural vehicle design.
- Skills: CAD Simplification, Dynamic Simulation, Altair MotionSolve, Vehicle Dynamics, Virtual Prototyping
Undergraduate Research Intern
Mechatronics Lab, IIT Delhi
Remote
- Utilizing Visual C# to develop and enhance the backend of RoboAnalyzer, a robotics analysis and visualization software.
- Implementing new features to improve simulation accuracy and system performance.
- Debugging and optimizing existing modules to ensure scalability and reliability.
- Collaborating with research mentors to align backend improvements with robotics research and use cases.
- Skills: Visual C#, Backend Development, Robotics Software, Debugging & Optimization
Undergraduate Research Intern
CMATER Lab, Jadavpur University, Kolkata
Remote
- SPAD Value Estimation: Formulated an ensemble approach combining CNN, DNN, XGBoost, and Random Forest with a linear regression meta-learner to predict chlorophyll (SPAD) values from rice leaf images. Achieved R² score of 0.7820, outperforming individual models. Published in COMSYS 2025 (Warsaw, Poland).
- Cassava Leaf Disease Classifier: Built a multi-backbone ensemble (ViT, ResNet-50, AlexNet, ConvNeXt, Swin-T) to classify 21,367 cassava leaf images across 5 disease categories. Achieved 87.03% validation accuracy and 0.7733 macro F1-score, improving results by 2–8% over individual backbones.
- Skills: Ensemble Learning, Deep Learning, CNN, DNN, XGBoost, Random Forest, Vision Transformer, ConvNeXt, PyTorch, TensorFlow/Keras
Undergraduate Research Intern
Robotics and Automation Laboratory, IIT Patna
Bihta, Bihar, India
- Developed transformation matrices between two points on free-form surfaces, computing rotation angles and translational vectors for toolpath generation.
- Implemented algorithms in MATLAB and Python for STL file rotation, point selection, and meshing-based coordinate extraction.
- Created codes for selecting points on STL files using meshing and applying computed rotation/translation for accurate orientation.
- Validated methodology on hemisphere and cube geometries, demonstrating accurate toolpath alignment and applicability to CNC machining.
- Skills: Computational Geometry, MATLAB, Python, Rotation and Translation Matrices, Euler Angles, 3D Meshing, CNC Toolpath Simulation
Undergraduate Research Intern
National Institute of Technology Durgapur
Durgapur, West Bengal, India
- Biped Robot Jumping Mechanism: Designed walking and jumping control strategies for a biped robot using MATLAB and Simulink. Developed simulation models for dynamic balance and ground reaction forces in collaboration with Mr. Debanuj Roy.
- UAV Tracking Controller: Developing and implementing a tracking sliding mode controller for unmanned aerial vehicle dynamics using MATLAB and Simulink. Simulating UAV trajectory tracking performance with robustness against external disturbances, in collaboration with Mr. Siddhartha Kundu.
- Eye Disease Classification: Applied deep learning and transfer learning models (DenseNet169, MaxViT-Tiny, EfficientNet-B0, ResNeXt-101) for eye disease classification. Achieved highest performance with DenseNet169, reaching 93.7% validation accuracy, 92.5% test accuracy, and 0.925 macro F1-score, outperforming other models by 2–10%.
- Skills: MATLAB, Simulink, Biped Robotics, UAV Control, Sliding Mode Control, Deep Learning, Transfer Learning, DenseNet, EfficientNet, ResNeXt, PyTorch