Sumesh Kalambettu Suresh
Hi there! I'm a Mechanical Engineering graduate from Carnegie Mellon University. I am passionate about Machine Learning and Computer Vision. During my time at CMU, I was a Research Assistant at Mechanical and AI lab under the guidance of Dr. Amir Barati Farimani and Francis Ogoke. I was also a Teaching Assistant at Language Technologies Institute for the course 11-785 Introduction to Deep Learning by Prof. Bhiksha Raj.
Prior to CMU, I completed my undergraduate in Mechanical Engineering from SRM Institute of Science and Technology, where I had the chance to work with Dr. Prabhu Sethuramalingam, Dr. A. Vijaya and Mr. A. C. Arun Raj.
Currently, I am a Machine Learning Research Engineer at FPrime AI working with Intel on a DARPA project, where I leverage deep learning for atomic timekeeping and digital signal processing.
Github LinkedIn E-mail
FPrime AI |
Language Technologies Institute |
Mechanical and AI Lab |
Carnegie Mellon University |
ABB India |
SRM Institute of Science and Technology |
Design and fabrication of soft robotics actuator for ABB IRB 360
Generative design for steering knuckle
Design of steering knuckle for CNC milling process
Design and manufacturing of freeform surface using VMC
2023
Master of Science in Mechanical Engineering - Applied Advance Study From Carnegie Mellon University.2022
Bachelors in Technology in Mechanical Engineering from SRM Institute of Science and Technology
Deep Learning based Optical Image Super-Resolution via Generative Diffusion Models for Layerwise in-situ LPBF Monitoring
Francis Ogoke, Sumesh Kalambettu Suresh, Jesse Adamczyk, Dan Bolintineanu, Anthony Garland, Michael Heiden, Amir Barati Farimani
Under review
Defects in Laser Powder Bed Fusion (L-PBF) limit its use in high-precision applications. Optical monitoring can detect defects but is costly and challenging to scale at high resolutions. We address this by training a conditional diffusion model to generate high-resolution build plate images from low-cost, low-resolution webcam images, capturing fine features and surface roughness. Model performance is evaluated using metrics like PSNR, SSIM, and wavelet covariance. We also leverage the Segment Anything foundation model to reconstruct 3D morphology and analyze surface roughness. Lastly, we test the frameworkâs zero-shot generalization to other part geometries using synthetic low-resolution data.
Paper | Code (Coming soon)
Design and development of universal soft robotic end effector for IRB 360 robot
K.S. Sumesh, V. Darshan, Dr. Prabhu Sethuramalingam, Dr. M. Uma
Bachelor's Thesis
The end effector, or gripper, is a critical component of robotic systems for tasks like grasping, carrying, and assembling objects. Soft robotic grippers, known for their flexibility in handling various shapes and materials, enhance adaptability in production environments. This study focuses on designing and analyzing soft robotic grippers for repeatability and high payload capacities. Machine learning (ML) is employed to optimize gripping forces based on the object type, ensuring precision and control. The gripper, designed for the IRB 360 flex picker robot, is tested in both virtual and experimental settings, achieving 94% accuracy in force prediction. Operating effectively at pressures between 1.4 and 2.8 bars with a maximum payload of 500 g, the gripper's angle and force are optimized using image processing and ML techniques.
paper | report
2024
Machine Learning Research Engineer FPrime AI2023
Teaching Assistant 11-785 Introduction to Deep Learning Fall 20232023
Research Assistant Mechanical and AI Lab2022
Robotics Intern ABB India (Discrete Automation and Robotics Division)2021
Manufacturing Lead Team 1.618 (Formula Hybrid Team of SRM Institute of Science and Technology)