Anand K Subramanian





github-iconAntixKlinkedin-iconanandkrishakmail-iconanandkrish894@gmail.compdf-iconResume

Summary

  I am a Machine Learning Engineer with 5 years of experience developing ML/DL models as well as conducting research on deep neural networks. I love working closely with hardware - developing DL models that are performant on the designated hardware and being Bayesian about it. I was a MEXT scholar during my Master's in JAIST, Japan, where my research was on spectral normalization for deep neural networks. Previously, I worked on embedded systems and signal processing. For one for my then projects, I had to learn statistical signal processing, which segued into my current interest in Machine Learning. Over the course of my professional career, I have been successful in applying Deep Learning for industrial anomaly detection, 3D PointCloud processing, Sim2Real, and multi-task learning.

Experience

Bodygram, Tokyo

Nov 2023 - Present

Machine Learning Engineer

Developing Deep-Learning solutions for accurate human body measurement and understanding.

#3D-Human-Understanding #Pytorch #SMPL #AWS

DeepX Inc, Tokyo

Feb 2022 - Nov 2023

Senior Computer Vision Engineer

Developed Deep-Learning solutions for automating heavy construction vehicles like excavators and trucks. Worked predominantly on 3D computer vision, and point-clouds. Tasks include object detection, segmentation, tracking, and Sim2Real.

#ROS2 #Pytorch/libtorch #OpenCV #ONNX #PCL #Sim2Real

NABLAS Inc, Tokyo

Oct 2019 - Jan 2022

Research Engineer

Built DL solutions for industrial anomaly detection with real-world constraints.
Developed a zero-shot anomaly detection solution and optimized it for deployment on an NVIDIA edge-device.

#Pytorch #OpenCV #ONNX #Docker

RIKEN AIP, Tokyo

Feb 2019 - May 2019

Visiting Researcher

Worked predominantly on Gaussian Processes - a variational technique for scaling them to a magnitude more data points.

#TensorFlow #GPy #Numerical Optimization

IIT Madras, India

Jun 2015 - Feb 2017

Project Associate

Main focus was on signal processing - giving lectures and tutorials on the ADSP-BF609 DSP chip.
Build the first prototype (now commercialized) for a low cost addressable fire alarm.

#Numerical Optimization #Signal Filtering #C++

Education

Japan Advanced Institute of Science and Technology (JAIST) 北陸先端科学技術大学院大学, Japan Oct 2017 - Dec 2019

Master's Degree in Informatics

Outstanding Student Award 2019
JAIST President's award 2019
Thesis: Mean Spectral Normalization for Deep Neural Networks.

SRM University, India Apr 2012 - Apr 2015

Bachelor's Degree in Mechatronics

CGPA - 8.9/10. Thesis: Autonomous robot swarm search using Alpha-Beta coordination. Link

Projects

PyTorch VAE

A collection of 18+ VAE implementations in PyTorch Lightning

PyTorch-Model Compare

A scalable way to compare the layer representations of neural networks

Mean Spectral Normalization

A plug-and-play alternative for Batch Normalization with fewer parameters for DNNs

Swarm Robot Project

Autonomous Robot Swarm for Searching using Alpha Beta Coordination

FCML

A collection of my notes on Machine Learning as a book

Curve Studio

A minimalist app for designing spline curves


Publications

Subramanian, Anand Krishnamoorthy, and Chong, Nak Young. "Mean Spectral Normalization of Deep Neural Networks for Embedded Automation." 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE). IEEE, 2019. DOI: 10.1109/CASE.2019.8842955

Subramanian, Anand Krishnamoorthy, and Girish Kumar. "Estimation of Optimum Robot Heading Using Savitzky-Golay and Kalman Filters." International Journal of Robotics and Automation. 1.2 (2016): 01-09. DOI: 10.37628/ijra.v1i2.25

© 2024 Anand K Subramanian License Design Built with Kutti