KEDAR
TATWAWADI
Contact: 47 Olmsted Rd, Stanford Phone: 669-238-1150 Mail: [email protected]
EDUCATION
Stanford University
• PhD in Electrical Engineering; Advisor: Prof. Tsachy Weissman [2015-present]
• MS in Electrical Engineering; CGPA – 4.13/4 [2015-17]
Indian Institute of Technology Bombay
• M.Tech. & B.Tech. in Electrical Engineering and Minor in Computer Science; CGPA – 9.58/10 [2009-14]
RESEARCH INTERESTS
I am broadly interested in data compression, machine learning, deep learning & its applications to the field of genomics
PUBLICATIONS
GTRAC: fast retrieval from compressed collections of genomic variants
K. Tatwawadi, M. Hernaez, I. Ochoa, T. Weissman in Bioinformatics, 2016
Compression of genomic sequencing reads via hash-based reordering: Algorithm and Analysis
S.Chandak, K.Tatwawadi, T.Weissman in Bioinformatics, 2017
MAJOR AWARDS AND SCHOLARSHIPS
• Awarded Numerical Technologies Founders Prize for achieving 2nd rank in Stanford PhD Qualifying Exam [2016]
• 3 time Gold Medalist at the International Olympiad on Astronomy and Astrophysics (IOAA) [2006-09]
• 3 time recipient of the Most Skillful Sky Observer Award (1 out of 200) at IOAA
• IIT Bombay Institute Prize for excellence in academics [2012]
• Awarded Gold Medal for Rank 6 at the Indian National Mathematics Olympiad [2008]
• Recipient of the KVPY Young Scientist Scholarship, Govt. of India [2007]
RESEARCH EXPERIENCE
Compression with Fast Random Access [2016]
Mikel Hernaez, Idoia Ochoa, Tsachy Weissman
• Motivation: Supporting efficient retrieval of any block of the compressed data, without decompressing the entire dataset
• Developed a compressor: GTRAC which supported 100x faster random access of genomic variant dataset
Compression of genomic sequencing reads [2016-17]
Shubham Chandak, Kedar Tatwawadi, Tsachy Weissman
• Motivation: Efficient compression of a collection of reads (substrings) of the DNA sequence, generated during the high
throughput genome sequencing
• Designed a genomic reads compressor which achieves 1.4-2x improvement by appropriate reordering of the reads
• Developed a systematic theoretical framework for analyzing the fundamental limits of reads compression
Deepzip: Lossless compression using recurrent networks [2017]
CS224n: Natural Language Processing using Deep Learning
• Designed DeepZip: a sequence compressor using Recurrent Neural Network (RNN) based prediction and Adaptive
Arithmetic coding, which achieves significant compression on data with long-range dependencies
• Performed a systematic information theoretical analysis of the fundamental limits on compression using RNNs
• Recognized as one of the outstanding course projects (10 out of 700)
ColorNet: Coloring B/W images using Deep Learning [2016]
CS231n: Convolutional Neural Networks for Visual Recognition
• Designed a convolutional neural-network based framework, ColorNet for colorizing a B/W image
• ColorNet was trained using Transfer Learning approach on the VGG16 network, using the ImageNet dataset
Compressed Sensing based imaging techniques in Radio Astronomy [2013-14]
IIT Bombay M.Tech. Thesis; Advisor: Prof. Sibi Raj Pillai
• Improved the signal processing operations of radio telescopes, by using compressed sensing algorithms to retrieve radio
image from its incomplete and noisy Fourier representation
• Developed heuristics to determine optimal parameters for various types of radio sources
INDUSTRIAL & TEACHING EXPERIENCE
• Ford Motors: Collaboration project on Compression of Video data [16-17]
• Goldman Sachs, India: Senior Analyst, Structured Products Group [Jun 2014 – Aug 15]
• IBM Research, India: Business Analytics and Optimization Group [2012]
• Teaching Assistant, Stanford: EE376a: Information Theory, EE376c: Universal Schemes in Information Theory
