All Projects

The true sign of intelligence is not knowledge but imagination. - Albert Einstein

Reinforcement Learning And Its Application in Smart Electricity Markets

August 16, 2019

  • Technology Stack – Python: OpenAI Gym, Tensorflow, Numpy, Scipy, Matplotlib
  • The project aims to model a smart market consisting of a single broker, which tries to minimise the difference between the profit and loss, and tries to replicate and incubate the dynamic substance of the market.
  • It uses the semi-supervised technique of Reinforcement learning, to model the actor, critic and the environment.
  • For model creation, development and application of algorithms, and simulation, we used OpenAI Gym.
  • We used the public data available at Independent Electricity System Operator (IESO). IESO contains various reports generated by the Ontario’s power grid systems, and provides detailed reports indicating the demand, supply, tariffs and all other parameters.

Machine Learning-based Optical Fine Alignment Tool

May 13, 2019

  • Technology Stack – Zemax OpticStudio, ZOS-API, Python: Numpy, Scipy, Scikit-learn
  • The project aims to model the variations in the Zernike Coefficients due to the misalignment of the optical components of the telescope, and then apply various Machine Learning algorithms to accurately predict the exact component having the misalignment, and the actual misalignment.
  • The project consists of two components: Generation of Dataset and Application of various Machine Learning algorithms.
  • The datasets consist of Standard Zernike Coefficients (37 coefficients) generated at the image plane, at all possible perturbations under the optical and mechanical tolerances defined previously, created using ZOS-API for Python.
  • The optical diagram of the system has been previously created using the Zemax OpticStudio, and used as a base to generate various perturbations using the API.
  • The dataset created, was preprocessed and standardized.
  • A Multi-class Random Forest Classifier was used to predict the exact component of the mis-alignment, and every class had its own Multi-Output Decision Tree Regressor, so as to predict the actual misalignment in all possible parameters.
  • This tool is proposed to be used in the Solar Ultraviolet Imaging Telescope (SUIT), one of the payloads of the Aditya-L1 mission of Indian Space Research Organization (ISRO)

LastMile

January 25, 2019

  • Technology Stack – Django, Solidity, Metamask, Javascript
  • Developed in 36 hours at Hack36, MNNIT Allahabad
  • The project aims to automate the process of the compensation provided to the passengers when flights are delayed/canceled, by leveraging the power of smart contracts.
  • As soon as the tracking system sends the details of the flights cancelled, the contract gets triggered and issues the meal, hotel or taxi vouchers to all the passengers of that flight.
  • The Airlines can use the developed API to drive the passenger experience.

Astronomical Time Series Analysis

January 10, 2019

  • Technology Stack – Python: Numpy, Scipy, Astropy, astroML, Tkinter
  • The project demonstrates Spectral Analysis and Forecasting performed on Time Series obtained from Astronomical Observations
  • The project employs the use of various Periodogram-based techniques like Fourier Transform, Lomb - Scargle Method, to estimate the Power Spectral Density at various frequencies.
  • The accurate forecasting of the data was done by estimating the various parameters of an ARIMA model, by the analysis of the Auto-Correlation Function (ACF) and the Partial Auto-Correlation Function (PACF) of the data.
  • The above-mentioned methods were employed on both evenly-sampled data (LIGO) as well as unevenly-sampled data (LINEAR)

SecTra

October 1, 2018

  • Technology Stack – Django, Kotlin, TensorFlow, AWS
  • Developed in 36 hours at Prototype, IIIT Allahabad
  • The project employs the use of Object Detection to identify the items not allowed in flight travels. It also differentiates between the items allowed in carry bag or check-in bag.
  • The airport security can use the software in assisting them to identify vulnerable items, which generally is a manual process.
  • Also, the project has an Android Application which assists the user in sorting his items into different bags based on the fight (Domestic/ International).
  • The project received 5th position at the Hackathon.

Text-To-Image Generation using Generative Adversial Networks

August 1, 2018

  • Technology Stack – Python: TensorFlow, Tensorlayer, Numpy, Scipy
  • The project employs the use of Generative Adversarial Networks (GAN) to convert an input text into corresponding images.
  • The model was trained on VGG Flowers and UCSD Birds dataset. Also, every image has 10 captions each, relating to corresponding image in the dataset.
  • The text was converted into feature vectors using LSTM Embeddings. They were concatenated with images and were feed on to the network.
  • The model used Adam Optimizer to optimize the cost function of the Generative Network.

Flawnkid

March 18, 2018

  • A kid - centric launcher that focuses on putting parental control on apps and time spent on apps by their kids, while gamifying the whole process to make it rewarding for kids as well.
  • Worked mainly on the launcher and the UI.

Caviar

March 1, 2018

  • Technology Stack – Django, PostgreSQL.
  • A web - based restaurant management system, which has User, Admin, and Staff interface.
  • User can select various food items, manage quantities and place the order for delivery.
  • Admin can view all orders of different customers, confirm the order and assign a delivery boy for the delivery boy for the item.
  • Staff (Delivery Boy) can view all the orders assigned for delivery, and update admin on completion.
  • Currently deployed on Heroku.

Aparoksha '18 Official Android Application

February 1, 2018

  • Technology Stack – Kotlin, Firebase (Realtime Database, Authentication, Analytics, Messaging), REST APIs, WaspDB
  • Used many open source libraries such as Retrofit, Moshi and Glide.
  • Fetches events info and latest updates, and presents to the users.
  • Employs usage of Android Architecture Components and Kotlin Coroutines.
  • Provides user with a QR code upon sign in, which is used then to register for the event.
  • App had around 500 downloads, and is available on Play Store.

Aparoksha '18 Campus Ambassador Application

December 1, 2017

  • Technology Stack – Kotlin, Firebase (Realtime Database, Storage, Authentication, Functions, Analytics, Messaging), REST APIs
  • The app is centered for a particular user, providing various features.
  • User can upload images for particular tasks, and points are credited appropriately.
  • A leaderboard is maintained to view top 10 users (with maximum points).
  • On reaching particular levels, scratch cards are awarded, which fetches more points.
  • The app is available on Google Play Store, with over 1000 downloads, all over the country.

Applet Communication

October 1, 2017

  • Technology Stack – Java (AWT, Swing)
  • The project demonstrated communication between two applets.
  • An applet displays an animation, whose parameters are controlled via second applet.
  • The project is implemented in two instances – Core AWT and AWT + Swing (Java GUI Frameworks).

Effervescence '17 Official Android Application

August 1, 2017

  • Technology Stack: Android, Kotlin, Firebase, REST APIs, JSON.
  • The App contained all data regarding events that occur during the fest.
  • Also displayed live updates regarding several events.
  • User can bookmark the event, and also set reminder for the events that he/she may be interested in.
  • Organizers can send notifications to all the users regarding update of events, using the organizers app.