Working on challenging problems and sizeable alternate data sources to draw meaningful insights for investors. Prominent topics that I have worked on include:- Anomaly detection for time series.- Time series characterisation (trend, correlation, peak, seasonality etc.).- Topic modeling (NLP).- Enterprise graph analysis for ranking startups, people, investors and universities. - Feed timeline for discovering new insights.# MachineLearning #DataScience # Investing #AlternateData #BigData

Worked on automatic assessment of music performances in the context of MOOC platforms. #machine learning | #music information retrieval | #digital signal processing

My thesis is titled "Computational Approaches for Melodic Description in Indian Art Music Corpora" supervised by Prof. Xavier Serra. This work was done as a part of the CompMusic project. The thesis aims to develop data-driven approaches for higher-level melodic description of Indian art music corpus by focusing on discovery and characterization of melodic patterns, and using them to perform tasks such as rāga recognition.For details visit: http://compmusic.upf.edu/node/304Teaching: - Teaching assistant for MOOC course ASPMA offered through Coursera for two consecutive years 2014-2015 - Teaching assistant for undergraduate course in UPF for three consecutive years 2013-2015.

Worked on automatic tonic identification of the lead singer from audio recordings of Indian art music. Masters dissertation can be obtained from: http://mtg.upf.edu/node/2595

Creating technology solutions for music education and infotainment. For more information visit: http://musicmuni.com/#Machine learning | #Data analytics | #Signal processing

Developed high precision guitar tuners for mobile apps and automatic assessment of guitar performance.

Developed a song version identification system capable of identifying different versions of a song, which are heavily remixed, sung in different language and even sometimes copied (plagiarism). The work mainly involved identifying and implementing relevant features and distance measures which suit this task in the context of Indian bollywood music. A intermediary step involved extracting signature phrase of a song (melodic hook) which itself brings out interesting applications.

Studied rhythm related aspect of light Indian classical music (for example songs with the tabla playing a certain tala), specifically automatic estimation of perceptual tempo and meter (double/triple/septuple etc) from the audio recordings. The research was guided by Prof. Preeti Rao in DAP Lab.Also as a part of this effort a rhythm toolbox was developed in C to facilitate rhythm related studies around Indian art music.

Worked on development of G3-FAX modems and voice enhancement devices (noise cancellation and acoustic echo cancellation). The work on G3-FAX modem involved developing transmitter and receiver for fax data sent on telephone lines (different channel codings) by taking care of the ITU-T guidelines and making system robust against channel degradations and jitters. Towards the later stages it also involved implementation of the V21 V48 fax protocols on DSP hardware.Developing voice enhancement units (noise cancellation and acoustic echo cancellation) was a quick learning experience of adaptive signal processing and speech related concepts. This project involved understanding the working of telephony in India and designing adaptive signal processing units accordingly to cancel the echo (both line and acoustic) due to the mismatch of the impedance or acoustic leakage in mobile handsets.

Worked on optimization of WMA decoder on ARM cortex A-8 processor.