Experience

Renude

• United States
• Manufacturing
• 1 - 100 Employee

• Computer Vision Engineer

  • Apr 2021 - Present

  • Developing a tiny object detection model for acne detection and localization using SSD, RPN, Region-based CNN (R-CNN), Faster R-CNN, YOLOv5, and transfer learning concepts. • Developing acne type and grade (binary and multi-class) classification utilising transfer learning-based modelling on pre-trained VGG, ResNet, and Inception models. • Developed ML pipelines to perform data pre-processing, spot training of object detector and classification models using Tensorflow and PyTorch estimators and docker, experimentation and hyperparameter tuning, saving model artefacts, and deploying the best fit model as an end point on the development environment using AWS SageMaker studio. • Built an algorithm in python to convert AWS Groundtruth JSON (for bounding box) to PascalVOC format. • Developed skin region of interest detection method to generate skin mask using OpenCV based image processing operations such as conversion to HSV color space, thresholding, smoothing, morphing. • Successfully segmented key facial regions such as the forehead, cheeks, chin, and nose area for profile and frontal-face images using fine-tuned CNN, DLIB based facial landmark detector and OpenCV utilities. • Setup and configured standard as well as custom labelling jobs for bounding box, single and multi-class labelling of skin lesions using on AWS SageMaker-Groundtruth and AWS Lambda. • Studied and explored various research papers and articles relevant to acne detection for varying image qualities.



ToriML

• Canada
• Software Development

• AI Engineer

  • Jan 2021 - Mar 2021

  ▪ Developed an open-source dynamic 3D-animated model (objects/characters) viewer and labeller with features such as zoom in/out and 720° rotation, using ThreeJS, PHP, Python, and DynamoDB. ▪ Mounted S3 bucket on EC2 using S3FS and integrated it with the web-app using S3-API to facilitate object availability to the web app. ▪ Assisted in research to develop a novel approach for AI-based models for real-time style transfer in Unity framework, utilising deep neural networks. ▪ Developed an open-source dynamic 3D-animated model (objects/characters) viewer and labeller with features such as zoom in/out and 720° rotation, using ThreeJS, PHP, Python, and DynamoDB. ▪ Mounted S3 bucket on EC2 using S3FS and integrated it with the web-app using S3-API to facilitate object availability to the web app. ▪ Assisted in research to develop a novel approach for AI-based models for real-time style transfer in Unity framework, utilising deep neural networks.



Virtusa

• India
• IT Services and IT Consulting
• 700 & Above Employee

• Technical Consultant

  • Jul 2018 - Sep 2020

  ▪ Develop and optimize data processing algorithms in the Hadoop system for ADSL/VDSL leased lines for various communication providers for BT.▪ Data pre-processing and analysis in Python for the data received from the CP.▪ Data visualization using Python to prepare dashboards for client review.▪ Handling missing values using Python Libraries to ensure data integrity and completeness.▪ Import and export data using Sqoop from HDFS to RDBMS and vice-versa.▪ Optimize Hive scripts to use HDFS efficiently by using various compression mechanisms.▪ Create Hive schemas using performance techniques such as partitioning.▪ Develop Oozie workflow jobs to execute Hive, Sqoop, Shell, and Map-Reduce actions.▪ Work in Agile based environment with daily scrum calls and coordination with team members.▪ Setup health check and algorithm monitoring Bash Scripts to raise alarms and auto retriggering of jobs in case of issues.▪ Set up process optimization script using Bash-Scripting to reduce human intervention and transfer data between remote systems such as automatic big data download script and error log readers.

• Technical Consultant

  • Oct 2016 - Jun 2018

  ▪ Developed, tested, and maintained an interactive web application for an inventory management system in a LAMP environment.▪ Worked in Agile based environment with daily scrum calls and coordination with team members.▪ Code integration, maintenance, and deployment using CI/CD tools▪ Performed server to server application migration involving technologies such as PHP, MySQL, Apache, and PostgreSQL▪ Setup Single Master to Multi Slave and Slave to Slave data replication for MySQL and PostgreSQL.▪ Provided a quick and efficient solution for bulk data movement (import/export) from the application via re- caching.▪ Optimized code to reduce the system load from several hours to few minutes.▪ Optimized data search time to reduce page loading time and availing results within seconds.▪ Provided strategic solutions for data replication issues and server load balancing.

• Senior Software Engineer

  • Oct 2015 - Sep 2016

  - Developed health-check scripts in bash and VB-Script to perform heartbeat checks for server health, availability and functionality.- Setup password less connectivity between Linux based servers using RSA token.- Achieved the tenets of ‘Right First Time (RFT) and Zero-Manual Intervention’ by automating time consuming and error prone manual efforts.- Created and maintained the self-learning and adaptive repository of issues, for service improvement.- Contributed in architectural redesign of legacy apps into new technologies.- Collaborated with Business users and offshore team in case of priority issues.- Collaborated business communications for service impacting issues.- Co-ordinated with offshore team and cross-functional teams to ensure that applications were properly configured tested and deployed.

• Software Engineer

  • Jul 2013 - Sep 2015

  - Performed software debugging and fixed a variety of code bugs.- Developed user interface using JSP.- Involved in writing complex SQL queries to query data.- Extensively used JDBC concepts to connect to, retrieve and update information in the database.- Performed instant progression of stuck faults raised by more than 300 users.- Identified issues from user perspective and investigated them.- Managed to Reduce stuck fault count by 30%, over a period of 3 months- Highlighted defects and code issues to respective technical teams.- Collaborated with Business users in case of priority issues.