{"id":1534,"date":"2013-08-13T01:26:18","date_gmt":"2013-08-13T00:26:18","guid":{"rendered":"http:\/\/pkmital.com\/home\/?page_id=1534"},"modified":"2023-05-31T22:38:48","modified_gmt":"2023-05-31T21:38:48","slug":"about","status":"publish","type":"page","link":"https:\/\/archive.pkmital.com\/","title":{"rendered":"About"},"content":{"rendered":"
\n

Parag Kumar Mital<\/h2>\n<\/div>\n
\n

Parag K. MITAL, Ph.D. (US) is an Indian American artist and interdisciplinary researcher working between computational arts and machine learning for nearly 20 years. He currently works as CTO and Head of research at HyperSurfaces working on AI audio research and at Never Before Heard Sounds as their Chief AI Scientist. His varied scientific background includes fields such as machine and deep learning, film cognition, eye-tracking studies, EEG and fMRI research. His artistic practice combines generative film experiences, augmented reality hallucinations, and expressive control of large audiovisual corpora, tackling questions of identity, memory, and the nature of perception. The balance between his scientific and arts practice allows both to reflect on each other: the science driving the theories, and the artwork re-defining the questions asked within the research.<\/p>\n

His work has been published and exhibited internationally including the Prix Ars Electronica, Walt Disney Concert Hall, ACM Multimedia, Victoria & Albert Museum, London\u2019s Science Museum, Oberhausen Short Film Festival, and the British Film Institute, and featured in press including BBC, NYTimes, FastCompany, and others. He has also taught at UCLA, University of Edinburgh, Goldsmiths, University of London, Dartmouth College, Srishti Institute of Art, Design and Technology, and California Institute of the Arts in both Undergraduate and Graduate levels in primarily computational arts applied courses focusing on machine learning applications. Finally, he is also a frequent collaborator with artists and cultural institutes such as Massive Attack, Sigur R\u00f3s, David Lynch, Google, Es Devlin, and Refik Anadol Studio.<\/p>\n

Feel free to contact me<\/a>.<\/p>\n

\"me\"<\/p>\n

\"scholar\"<\/a>\"github\"<\/a>\"twitter\"<\/a>\"flickr\"<\/a>\"vimeo\"<\/a><\/p>\n

Current Work<\/h2>\n

Founder<\/strong> (2010-current) The Garden in the Machine, Inc., Los Angeles, CA, U.S.A.
\nComputational arts practices incorporating applied machine learning research, generative film, image, and sound practices, and working with artists and cultural institutes such as Massive Attack, Sigur R\u00f3s, David Lynch, Google, Es Devlin, Refik Anadol Studio, LA Phil, Walt Disney Concert Hall, Boiler Room, XL Recordings, and others.<\/p>\n

CTO\/Head of Research<\/strong> (2017-current) HyperSurfaces<\/a>, Mogees, Ltd.<\/a>, London, U.K.
\nHyperSurfaces is a low cost, low power, privacy preserving technology for understanding of vibrations on objects and materials of any shape or size, bringing life to passive objects around us and merging the physical and data worlds.<\/p>\n

Chief AI Scientist<\/strong> (2023-current) Never Before Heard Sounds<\/a>, New York, U.S.A.
\nAI powered music studio in the browser. Leading the research and development of AI tools for musical content creation, composition, arrangement, and synthesis.<\/p>\n

Adjunct Faculty<\/strong> (2020-current) UCLA<\/a>, Los Angeles, CA, U.S.A.
\nCultural Appropriation with Machine Learning is a special topics course taught in 2020 and soon again in Fall 2023 where students of the Design and Media Arts program are taught both a critical framing for approaching generative arts as well as how to integrate various tools into their own practice.<\/p>\n

Previous Work Experience<\/h2>\n

Creative Technologist<\/strong> (2019-2020) Artists and Machine Intelligence, Google, Inc.<\/a>, Los Angeles, CA, U.S.A.
\nWorked with international artists, including Martine Syms, Anna Ridler, Allison Parrish, and Paola Torres N\u00fa\u00f1ez del Prado, to help integrate machine learning techniques into their artistic practice.<\/p>\n

Director of Machine Intelligence \/ Senior Research Scientist<\/strong> (2016-2018) Kadenze, Inc.<\/a>, Valencia, CA, U.S.A.
\nBuilt bespoke ML\/DL pipeline using Python, TensorFlow, Ruby on Rails, and ELK stack for recommendation, personalization, and search of various media sources on Kadenze, Inc. Setup backend using CloudFormation, ECS, ELK stack and various frontend visualizations of trained models on Kadenze data using Kibana, Bokeh, D3.js, and Python. Built bespoke ML\/DL solutions to enable analysis and auto-grading of images, sound, and code. Built, taught, and continue to deliver Kadenze\u2019s most successful course on Deep Learning, Creative Applications of Deep Learning w\/ TensorFlow, and fostered and built collaborations with Google Brain and Nvidia (including usage of their HPC cluster) to partner on the course.<\/p>\n

Artist Fellow<\/strong> (2016-2018) California Institute of the Arts<\/a>, Valencia, CA, U.S.A.
\nMentorship and development of arts practice with BFA, MFA students in the Music Technology and Intelligent Interaction Design<\/a> program. Teaching of “Audiovisual Signal Processing” and “Augmented Sound”.<\/p>\n

Freelance Consultant<\/strong> (2017) Google and Boiler Room<\/a>, London, U.K.
\nDeveloped bespoke neural audio synthesis algorithm for public launch of the Google Pixel 2 phone, including backend for website which served pre-trained neural audio model for stylization and synthesis and live processing of voices. Built with node.js, Redis, TensorFlow, and Google Cloud.<\/p>\n

Freelance Consultant<\/strong> (2017) Google Arts and Culture\/Google Cultural Institute<\/a>, Paris, FR
\nPorted a model to TensorFlow to reproduce results of a Inception-based deep convolutional neural network written in another framework.<\/p>\n

Senior Research Scientist<\/strong> (2015) Firef.ly Experience Ltd.<\/a>, London, U.K.
\nMachine learning and signal processing of user behavior and activity patterns from GPS and smartphone motion data.
\nMongoDB cluster computing; Mapbox; Python; Objective-C, Swift; Machine learning; Mobile signal processing.<\/p>\n

Visiting Researcher<\/strong> (2015) at the Mixed Reality Lab Studios<\/a> at the University of Southern California<\/a>, Los Angeles, CA
\nAugmented reality; Unity 5; Procedural audiovisual synthesis.<\/p>\n

Post-Doctoral Research Associate<\/strong> (2014-2015) at the Bregman Media Labs<\/a> at Dartmouth College<\/a>, Hanover, NH
\nExploring feature learning in audiovisual data, fMRI coding of musical and audiovisual stimuli during experienced and imagined settings, and sound and image synthesis techniques. Designed experiment for fMRI and behavioral data, collected data using 3T fMRI and PyschoPy, wrote custom pre-processing of data using AFNI\/SUMA\/Freesurfer using the Dartmouth Discovery supercomputing cluster, and developed methods using Univariate and Multivariate methods including Hyperalignment measures using PyMVPA. Principal Investigator: Michael Casey<\/p>\n

Research Assistant<\/strong> (2011) London Knowledge Lab<\/a>, Institute of Education, London, U.K.
\nECHOES<\/a> is a technology-enhanced learning environment where 5-to-7-year-old children on the Autism Spectrum and their typically developing peers can explore and improve social and communicative skills through interacting and collaborating with virtual characters (agents) and digital objects. ECHOES provides developmentally appropriate goals and methods of intervention that are meaningful to the individual child, and prioritises communicative skills such as joint attention. Wrote custom computer vision code for calibrating behavioral measures of attention within a large format touchscreen television. Funded by the EPSRC. Principal Investigators: Oliver Lemon and Kaska Porayska-Pomsta<\/p>\n

Research Assistant<\/strong> (2008-2010) John M. Henderson\u2019s Visual Cognition Lab, University of Edinburgh
\nInvestigating dynamic scene perception through computational models of eye-movements, low-level static and temporal visual features, film composition, and object and scene semantics. Wrote custom code for processing large corpus of audiovisual data, correlating the data with behavioral measures from a large collection of human subject eye-movements, and applied pattern recognition and signal processing techniques to infer the contribution of auditory and visual features and their interaction within different tasks and film editing styles. The DIEM Project<\/a>. Funded by the Leverhulme Trust and ESRC. Principal Investigator: John M. Henderson<\/a><\/p>\n

Education<\/h2>\n

Ph.D. (2014) Arts and Computational Technologies<\/strong> Goldsmiths, University of London.
\nThesis: Computational Audiovisual Scene Synthesis<\/a>
\nThis thesis attempts to open a dialogue around fundamental questions of perception such as: how do we represent our ongoing auditory or visual perception of the world using our brain; what could these representations explain and not explain; and how can these representations eventually be modeled by computers?<\/p>\n

M.Sc. (2008) Artificial Intelligence<\/strong>: Intelligent Robotics, University of Edinburgh
\nB.Sc. (2007) Computer and Information Sciences<\/strong>, University of Delaware<\/p>\n

Publications<\/h2>\n
<\/p>\n
\n

Parag K Mital. Time Domain Neural Audio Style Transfer<\/strong> Neural Information Processing Systems Conference 2017 (NIPS2017), https:\/\/arxiv.org\/abs\/1711.11160, December 3 – 9, 2017<\/p>\n

[github]<\/a> [arxiv]<\/a><\/p>\n<\/div>\n<\/div>\n

<\/p>\n
Christian Frisson, Nicolas Riche, Antoine Coutrot, Charles-Alexandre Delestage, St\u00e9phane Dupont, Onur Ferhat, Nathalie Guyader, Sidi Ahmed Mahmoudi, Matei Mancas, Parag K Mital, Alicia Prieto Ech\u00e1niz, Fran\u00e7ois Rocca, Alexis Rochette, Willy Yvart. Auracle: how are salient cues situated in audiovisual content?<\/strong> eNTERFACE 2014, Bilbao, Spain, June 9 – July 4, 2014.<\/p>\n

[online]<\/a> [pdf]<\/a><\/p>\n<\/div>\n<\/div>\n

<\/p>\n
\nParag K. Mital, Jessica Thompson, Michael Casey. How Humans Hear and Imagine Musical Scales: Decoding Absolute and Relative Pitch with fMRI.<\/strong> CCN 2014, Dartmouth College, Hanover, NH, USA, August 25-26, 2014.<\/p>\n<\/div>\n<\/div>\n