ASHUTOSH RAJ

Computer Vision - Research Engineer

adaht1@r.postjobfree.com

+91-799*******/847-***-****

Whitefield

Bangalore - 560037

Karnataka

linkedin.com/in/ashutoshraj

github.com/ashutoshraj

PROFILE SUMMARY Hands-on experienced professional involved in the several POC and products. Worked on a machine vision platform that requires optimized pipeline planning, design thinking, and deep learning/image processing algorithms for cloud environment configuration and hardware deployments. PROFESSIONAL WORK EXPERIENCE PATENTS

Tata Consultancy Services Limited - Corporate

Research and Innovation Labs

January 2017 - Present

● Worked on ML-based pattern recognition to

detect, segment and recognize human faces

and objects from a still images and videos.

● Created an algorithm for digital video

stabilization and image alignment to reduce

camera ego-motion from the video feed of

moving objects like drones.

● Improved customer experience for security

systems using face authentication algorithms.

● Worked on prediction and anomaly detection

algorithms from business critical time series

data to generate visual insights.

● Created CI/CD pipeline with automated

performance testing and memory profiling for

ML-based algorithms.

ACHIEVEMENTS

● Tata Innovista-2018 Finalist Team for Warehouse

Inventory Reconciliation Project.

● Received appreciation from the clients for Asset Inspection & Geyser Monitoring System.

● Won AIP Anchor Award by Academic Interface,

TCS Internal Programme for conducting training

sessions on Image Processing and Deep Learning

in various colleges in Kerala.

LEARNINGS & CERTIFICATIONS

● Secure & Private AI Scholarship, Udacity, 2019

● Machine Learning A-Z TM: Hands-On Python & R

In Data Science, Udemy, 2018

● edX Honor Code Certificate for Introduction to

Computer Science and Programming Using

Python, Edx, 2015

● Fundamentals of Digital Image and Video

Processing, Coursera, 2013

● Intro to Deep Learning with PyTorch, Udacity,

2019

● CS231n: Convolution Neural Network for Visual

Recognition, Stanford University, 2017

● System and method for vegetation monitoring

on power lines - (IDF-104****-***)

● Cognitive document verification and

information retrieval RPA - (IDF-920111-035)

SKILLS

Linear Algebra, Information Theory, Semantic

Segmentation, Object Detection, Image Alignment,

Core Image Processing, Point Cloud Processing,

Design Patterns, Actor System(Thespian), Differential Privacy and Federated Learning.

LIBRARIES & FRAMEWORKS

Image Processing:OpenCV, Scikit-image, PIL/Pillow

Computation Libraries:Numpy, Pandas, Open3D,

gprof.

Machine Learning:Scikit-learn, Fast.ai, Pytorch,

Tensorboard, Keras(Tensorflow), Tensorflow Serving, ONNX, t-SNE, PCA, Tensor-RT.

IDE:Pycharm, Jupyter Notebook, Vi/Nano, Sublime

Text.

Databases: MongoDB, PostgreSQL

Programming: Python, C++ and Shell

(Bash/Terminal).

Real-Time Communication: WebSocket.

Virtualization: VirtualBox, Nivida-docker.

Cloud Deployment: AWS Sagemaker, S3, EC2,

Lambda.

Software Versioning: Git.

INTERNSHIPS

Bharat Electronics Limited

July 2015 - August 2015

Completed training on "RADAR TRANSMITTER" with

excellent performance.

Sapro Imaging & Robotics

September 2013 - November 2013

Developed & deployed the project on Video

Surveillance and SpyBot.

Github link for Video Surveillance:

https://github.com/ashutoshraj/CameraSurveillance

PROJECTS

Mines Geyser-Outburst Monitoring System

Jan 2019 - Current

Feature-based analytics generation and predictions of an outburst of Geyser Hotspot in mines area using visual monitoring systems.

Based upon the previous data analysis of steam volume, originating geyser pores and several other parameters. Designed algorithm for finding n number of clusters without defining beforehand based on image processing. Able to detect geyser originating pores with 92.8% accuracy.

Created automatic Day/Night detection Algorithm for a large volume of videos.

Implemented video stabilization algorithm such as auto exposure, feature-based, detection based.

Generated plots for visualization of relevant parameters using the t-SNE and PCA algorithm.

Monitoring Vegetation Encroachment with Powerlines Nov 2018 - Mar 2019

Data is collected with LIDAR and analyzed to detect diseased trees that could fall and hit a powerline (fall ins) or trees that are encroaching on powerlines (grow-ins). Developed statistical models and machine learning

approaches for the analysis of point cloud data.

Every structural inferencing of Pole Detection was achieved using only LIDAR data.

Developed an algorithm for a 3D histogram upon the specified plane with varying distributions.

For the detection of trees and their species, RGB Images were involved.

Cognitive Document Verification And Information

Retrieval - RPA

Jan 2017 - Oct 2018

This is an internal project for the Human Resource team. This system goes through all the documents of new joiners to validate whether he/she has submitted all the required documents (Govt., Legal & Educational Docs) or not. The system uses robotic process automation to download all the docs and updates it on the website.

Developed code for automatic download and upload of documents through Desktop GUI Interface.

Used selenium, OCR, Image Processing and python

automation for automating the whole pipeline.

Signature Verification Without CNN

Jan 2018 - Mar 2018

Designed a lightweight python application using core image processing concepts and libraries.

Asset Inspection for Transmission Towers

Jul 2018 - Current

Intelligent predictive maintenance of Power

Transmission Tower using drones visual feed.

After data acquisition, we run our data processing for generating insights & analytics for various anomalies detection such as cracked insulators, flashed dampers, damaged conductors, etc.

Developed custom Dynamic-UNET architecture for

pixel-level (semantic) segmentation of insulators from the background.

Developed and deployed this project onto the AWS

infrastructure API based deployment.

Warehouse Inventory Reconciliation Using Drones

Mar 2018 - Oct 2018

Reconciling the whole warehouse inventories using

drone video feed.

Inventory Identification & Reconciliation involves barcode/ QR code

recognition, Text Recognition, Pallet Region Detection. Developed code for localizing the labels on the

inventory and implemented an actor system for

multiprocessing of the barcode decoding module.

Used several design patterns for designing the pipeline such as logging system, database wrapper.

Face Authentication System For Airlines Customers, Oct 2017 - Mar 2018

It involves Image Processing(Depth & Color stream) and Deep Learning algorithms to the differentiation between Photo/Video or Human Face.

After human verification, it compares the customer face with the Passport ID.

This project was deployed on the AWS server and

having a web application interface running web

sockets.

Intelligent Character Recognition System

Mar 2017 - Aug 2017

Developed and deployed the Optical Character

Recognition model on our own database with up to

78.2% accuracy.

The intelligent part comes while segmenting the type of data from different fields.

Developed code for the extraction of letters from the predefined physical form.

Created a simple desktop GUI application for the

same.

Details about this project are on Github

EDUCATION

Electronics & Communication, Bachelor Of Engineering & Technology, Graduated, July 2016 Krishna Institute Of Engineering Technology

Dr. A.P.J. Abdul Kalam Technical University – Marks 72%, Division I Lucknow, Uttar Pradesh

DECLARATION

I, Ashutosh Raj, hereby declare that the information contained herein is true and correct to the best of my knowledge and belief.

Ashutosh Raj

Bangalore, Karnataka

REFERENCES

**References available upon request.

Contact this candidate