I'm a software developer. The development of the projects I do revolve around Data Science, Full Stack Web, Desktop and Android.
Contact Me
You can reach out to me at:
me@mayankgupta.in
Startup India
Product of Smart India Hackathon 2017, Startup India is an Android Application, new and improved version of already existing app from Ministry of Commerce. An organisation can register as a startup, registered startup can check for their approval and registration, new user can view detail about registering a startup with ministry of commerce.
Spying application. Logs all the keystrokes, captures screenshot and sends them to your google drive folder every minute. Parental monitoring at its best.
Simple Android application to get a received SMS from specified sender and send it to the specified Web API. It automatically detects when the message is received from that sender, pretty much in the same way how applications send OTP via SMS and autofill it in the application by reading the SMS themselves.
Fully functional Inventory and accounts management application for Medicine Shops, may it be wholesale firm or a simple retail shop. The software has registration methods for both. Also, internal functionalities are also switched if login mode is changed (Wholesale/Retail)
Alexa Skill to order a product, retrieve updates on an already existing order, getting all orders, getting product information. Basically accessing E-commerce website with an ease of use.
This came out as a product of a Hackathon held at Cyber Group India pvt. ltd.
I mostly worked on interaction model which accesses the E-commerce web API to transfer all the interaction(to and fro) whilst customer interaction.
CCExtractor was selected as an organisation in GSOC’17. I made a GUI in C (using Nuklear library) for their terminal based application (Closed Captions Extractor)
This is an out of Company project I made just to learn tensorflow.js, but it came out pretty good. This application recognises faces on browser. That means, once the website is loaded and you’ve trained the software with your images, you can even plug off the internet and the facial recognition would work as it should.
I made this project completely from scratch, though I used premade python models for facial recognition and converted them to tensorflow.js compatible models.
I made this project as a starter to predict sentiment for company’s client’s data having phone call recordings converted to text. They wanted to know if the conversation was going in the right direction or not and also the intensity in a number so that they could reroute the call to a more suitable customer executive.
Though the final project is a part of much bigger project and was trained on data of different genres and not just movie reviews. However, I started out with Movie Reviews, made a python model using Neural networks. I made two to compare which technique works best.
I converted them to tensorflow.js compatible model, loaded them to browser, added D3 visualisation for comparison between two models and viola. I had real time comparison of a sentiment prediction model along with prediction confidence.