Shikhar Chauhan

Developing an AI driven Holistic Fitness Platform that helps users make fitness a habit.

My main responsibilities include development of Machine Learning models and API.

• Researching and Developing Machine Learning algorithms to develop holistic Diet plans and Workout plans for users based on their lifestyles and bodies using PyTorch
• End to End integration of Machine Learning Models with Backend
• Developing APIs to be consumed by Web and Mobile Devices using Django, Celery, MySQL, Redis and RabbitMq
• Writing tests and managing deployments using Pytest and Docker
• Integrating tools for monitoring performance and health of APIs

Facebook Open Source Mentorship Program is a 6 week program where I worked on open source projects under a Facebook Engineer.

I worked on Facebook Research's Visdom and Pytorch's Vision

• Visdom is a flexible tool for creating, organizing, and sharing visualizations of live, rich data. I worked on adding a readonly mode to the visdom server. https://github.com/facebookresearch/visdom/pull/319
• Vision is a pytorch module that consists of popular datasets, model architectures, and common image transformations for computer vision. I worked on adding CPU based gradient for ROIAlign. https://github.com/pytorch/vision/pull/504

Worked as freelance developer to develop algorithms and web services

• Worked on Diet plan generation algorithm
• Evolved the existing monolithic api of 98fit into micro services that can be leveraged across web and native android applications using Django , Redis and Nginx
• Integration of site analytics

Worked as a Backend Developer and Machine Learning Engineer with dedicated effort towards developing intelligent Chatbots

• Developed a prototypal Chatbot Framework built upon Recurrent Neural Networks and Non Deterministic Finite Automatons
• Developed Custom Data Containers to facilitate the development of intelligent Chatbot
• Developed APIs for interaction with Chatbot

Worked extensively on providing streamlined experience to users by developing intuitive UXs coupled with robust and low latency APIs.

• Database Design and Implementation
• Queue services based on Celery, Rabbitmq, Flask and SQLAlchemy for sending mails and managing SMSs
• Discussion Forums as single page applications using BackboneJS
• Discussion Forums Caching using Redis
• User Profiles as single page applications using BackboneJS
• Backend REST APIs for Discussion Forums and User Profiles using Flask
• Search APIs using NodeJS , ExpressJS and MongoDB
• Web Scrapers to find unexplored colleges and new information

Worked on Dashboard for the leaf wearables which provided realtime updates for the connected devices and response teams.

• Dashboard showed real time data incoming through websockets
• Implemented keyword based search to find relevant information from the dashboard

YCombinator Startup School is a MOOC taught by YC founders. Lessons included increasing user traction, revenue models, and scaling.

Project Forrest was selected as one of the startups.

Forrest is a platform where developers share programming resources they swear-by, organized in easy-to-follow paths.

Lego is a Python library that automatically assigns arguments passed to initializer to the object while creating an object of a class.

Philter is a messenger chat bot that manipulates images and videos using Artistic Style Transfer. It was built at Finals of Facebook Hackathon , 2016 , Facebook HQ, Menlo Park

You sign into it, it gives your own page and all your tweets with #meList are compiled to your webpage. All this with the asynchronicity of JS and the sweetness of Promises.

Mason was the web based lesson editor and builder for project MILE.

Built on BackboneJS, the application parsed the JS objects to YAML and made them available as a downloadable file. All this without using a backend.

MILE aimed to be a project which aimed to make learning not only discrete but visually and interactively stimulating as well.

It parses YAML based lesson plans to emulate lessons in MATLAB shell.