Repurposing Excel for CFO's office.

Areas of work: Machine Learning, Data Mining, and Natural Language Processing.1) Organizational data split across multiple data stores like RDBMS, Hadoop, excel files, csv files, Salesforce, CRM etc., makes it impossible to get a unified view of data. Auto Data Modeller (ADM) discovers a unified data schema by analyzing data from all data stores. I have proposed, prototyped, and developed the ADM, which powers DataRPM.2) Natural Language Interface to DB (NLIDB), prototyping… Show more Areas of work: Machine Learning, Data Mining, and Natural Language Processing.1) Organizational data split across multiple data stores like RDBMS, Hadoop, excel files, csv files, Salesforce, CRM etc., makes it impossible to get a unified view of data. Auto Data Modeller (ADM) discovers a unified data schema by analyzing data from all data stores. I have proposed, prototyped, and developed the ADM, which powers DataRPM.2) Natural Language Interface to DB (NLIDB), prototyping this system for querying data bases in natural language - english. Show less

Headed research team of a media house which is into digitization of historical news paper and journals, media monitoring, mobile application development for content publishers, and publication services.Areas of work: Information Retrieval, Machine Learning, Natural Language Processing, Mobile Applications.1) Near-real time indexer - given sereal hundreds of thousands of keywords, this system finds occurrences of all keywords in a document instantaneously. Keyword, phrase… Show more Headed research team of a media house which is into digitization of historical news paper and journals, media monitoring, mobile application development for content publishers, and publication services.Areas of work: Information Retrieval, Machine Learning, Natural Language Processing, Mobile Applications.1) Near-real time indexer - given sereal hundreds of thousands of keywords, this system finds occurrences of all keywords in a document instantaneously. Keyword, phrase, proximity, limited wild card searches are supported by this indexer. It also supports boolean expressions on keywords. This system is being used for media monitoring. 2) OCR error correction: auto corrects non-word errros in OCRd documents using machine learning techniques. This system is used in document digitization practice. 3) Epub3 reader library: a full featured JS library for rendering ePub3. 4) Comic Book Composer: A semi automatic system for converting a bunch of images into a comic book. Used by "Amar Chitra Katha" - a leading comic and story book publisher in India, for creating comic books from their image repository.5) Interactive eBook Composer: a self service web application for creating interactive EPubs with support for multimedia content. Show less

As an individual contributor in platforms team built: A) a media storage and delivery system - very similar to S3 much before Amazon came out with B) a single sign-on system C) founding engineer of travel.bixee.com - now goibibo.com D) friend recommendation system for IBIBO social network.

Built a hybrid content districution system which falls back to HTTP protocol in case file download with P2P protocol (Bittorrent) fails. Also worked on backend service implementation, search personalization, user and application analytics.