Responsibilities :

• Developing Language Models for Non-English languages to provide Conversational AI products.

• Working closely with the NLU research team in collaboration with the MILA Lab, Montreal.

  • Technologies used: Tensorflow, Python, Docker.
    

Responsibilities :

• Designed and Developed an end-to-end Deep-Learning based computer vision pipeline to remove shadows from images caused due to varying illumination conditions. The Shadow Removal algorithm I developed has improved the company’s existing algorithms performance by 95% and is ready for deployment in production.
• Researched and Developed Domain Adaption algorithms using Generative Adversarial Networks (GAN’s) for the Shadow-Removal project.
• Worked with an experienced team of Senior Software Engineers that followed industry-proven disciplined agile methodology for development.

• Project management using tools like JIRA to create, assign, and submit tasks, manage task descriptions and task timelines.

  • Technologies used: PyTorch, Python, OpenCV, Docker.
    

Responsibilities:

• Developing novel algorithms for Real-Time Semantic Segmentation on Low Power Devices.
• Developing algorithms for domain adaptation of semantic segmentation of hands using Generative Adversarial Networks (GANs).

• Developed and open-sourced a large scale dataset containing 200k annotated depth images and deep-learning architectures for hand segmentation.

  • More information on this project page
    

• Administration of the database server and part of a team that involved in the design of physical database schema, data modeling, and performance tuning. I was part of a research team working on building an Explosive Recommender System, that helps optimize blasting operation to attain production efficiency in coal mines.