Wikipedia has been facing severe backlog in its new page review process due to tremendous number of new pages, not enough reviewers and no mechanism to match expertise of reviewers to pages for quick review. This project demonstrates how this problem can be tackled predicting high level topics for new drafts on Wikipedia where the topic space is derived from an organic ontology defined by Wikipedians themselves.

The huge distributed codebase of Microsoft has created the need to improve development efforts using automated methods to save development time. Project sankie is a step towards that directions involving components like automatic code review suggestions, online root cause evaluation, _changelist risk prediction, etc.

Open source contributions to the core mediawiki engine that runs Wikipedia and the mobile team of Wikimedia Foundation that maintains Wikipedia on mobile.

Structured knowledge bases are becoming increasingly popular with for search engines. A common issue faced by search engines when dealing with type like relations like “person-profession” pair is that of ranking them based on their order of relevance. This undergraduate thesis project explores supervised methods for ranking relations in Web Search, considering preson-profession and person-nationality relations in particular.

This project demonstrates the usage of CNNs for question answering on the WikiQA dataset. The structure of the network is provided with some experimental results. This was a partial work done at University.

A collection of projects done at IIT Patna related to the artificial intelligence course. Includes a maze solver using the A* star algorithm, part of speech tagger using neural networks, part of speech tagger using Hidden markov models and a machine translation model from english to hindi. All algorithms except neural networks are implemented from scratch.

Cancer detection by looking calcifications in mammograms by doctors is a manual process that is prone to error given the size of calcifications. This digital image processing project attempts to make this process easier by isolating calcifications in the mammogram using image processing concepts. Implemented using MATLAB scripts.

Google Summer of Code 2015 project on WikidataPageBanner exetension. This project implements an extension for Wikivoyages which use a panoramic banner on each page. The extension makes this banner integration easier and connects it to Wikidata for picking up banners from there. Another important contribution of the project is towards making these banners mobile compatible which was an important part for Wikivoyagers.

A software engineering project to make teaching graph algorithms easier through the use of simulated illustrations of graph algorithms. The core model includes an abstract graph simulator built on d3 using which individual graph algorithms can be simulated with minimal code additions. Customizations include coloring the graph, controlling speed and manual state travel of the graph.