Default profile banner
VC

Vishnu Chand M. S.

@vishnuchandms

Machine Learning Engineer at Hippo Video

Chennai

https://www.linkedin.com/in/vi89

Hippo VideoPondicherry Engineering College

Vishnu Chand M. S. is an experienced professional with over 7 years in designing and deploying deep learning and computer vision applications. His expertise spans the entire development lifecycle, from research to deployment, including creating hybrid solutions to enhance deep-learning accuracy. He has experience developing and deploying solutions on AWS and holds a patent for dynamic multi-camera tracking.

Experience

Machine Learning Engineer

Hippo Video

•Invalid Date - Present•Chennai

Project (text2face): Goal was to generate video of a person speaking a transcript. Solution involved two stages: text2voice (using Google TTS and deep normalizing flow) and voice2face (using wav2lip GAN). Project (virtual background): Developed a cloud-based solution for replacing recorded video backgrounds using HRNet based semantic segmentation and FFMPEG. Project (virtual background speed improvement): Optimized inference speed by experimenting with optimal batch size, implementing thread queue, and using latent vector based skip inference.

Machine Learning Engineer

TCS

•Invalid Date - Invalid Date•Bangalore

Project: Tracking and counting the number of vehicles in Drive-Thru Canopies. Goal was to detect and track vehicles in drive-thru canopies, handling challenges like missing detection, rain, and static objects. Solution involved implementing a motion based confidence boosting algorithm, temporal edge suppression, and a novel vehicle tracking algorithm based on background and static occlusion modeling.

Education

Pondicherry Engineering College

Master Of Technology

Electronics & Communication

Jan 2014•Grade: 85.2% (distinction)

Specialization in Electronics and Communication (ECE).

Pondicherry Engineering College

Bachelor Of Engineering

Electronics & Communication

Jan 2011•Grade: 74.04%

Licenses & Certifications

Patent (11069072) for Dynamic multi-camera tracking of moving objects in motion streams

• No expiration

Credential ID: 11069072

Skills

Pytorch
Tensorflow 2.0
OpenCV
Numpy
Albumentations
Boto
pyDub
curl
ffmpeg
Ubuntu
aws EC2
conda/pip
pyCharm
Docker
wandb
KNN
K-Means
GMM
CNN
Semantic Segmentation
Copy-Paste Augmentation
Normalizing flows
Affine coupling
Online Hard Example Mining (OHEM)