Gaurav Ragtah

- Identifying genres of songs is an important problem in music information retrieval. - Used different machine learning techniques - Gaussian Mixture Modeling, Multi-class SVM, sparse representation classification to classify songs on the basis of their genre. - Used leave-one-out cross validation for evaluation and compared the validation rate across classifiers.

Research at Blei Lab exploring collaborative filtering from noisy GPS check-ins. Developed a modified LDA approach to jointly model user preferences and sensor noise, enabling more accurate inference of latent preferences and personalized place recommendations.

Built a web-based skill testing service POC that seeks to eliminate the subjectivity of resumes and false positives and negatives in the screening/hiring process. We provide relevant, standardized skill tests, allow independent, approved test-creators to auto-conduct their tests, and provide sophisticated candidate and test analytics. The website was implemented using the MEAN stack, and we worked on the business model and disruptive edge, drew up financial projections, performed competitive analysis, iterated on product design+dev, and created a full-fledged business plan over the course of the semester. Our work received the top grade for the Principles of Innovation and Entrepreneurship course at Columbia University.

Cloud 9 is an end-to-end dating application that helps sets up potential dates with matches through Tinder, uses NLP against the chat text to infer the match’s interests, preferences, time availability, etc. and helps schedule a date night at a great venue at a perfect time by leveraging Yelp and Google Calendar APIs, and calls an Uber for you when it’s time to leave for the date.

Graduate Teaching Assistant Fellow

Dean's Award for Academic Excellence (2010–2013)