CV

Resume

Education

• M.S. in CS, University of Colorado Boulder, 2020 (expected)
• B.E. in CS, University of Mumbai, 2013 - 2017.

Work experience

• Graduate Research Assistant - Institute of Cognitive Science: May 2019 - Present
  • My research lies at the intersection of Cognitive Science and Deep Learning for Multi-Label classification of fNIRS data to predict cognitive workload.
  • Day-to-day tasks involve getting my hands dirty with:
    1. Data Synthesis using SMOTE, Tomek links and ENN.
    2. Deep Learning with 3DCNNs, LSTMs (not mutually exclusive).
    3. Investigating model transfer across different fNIRS devices and Subjects.
    4. Investigating learned features using LIME and SHAP and deconvolution.
    5. Tons of reading and documentation.
    6. Publication in SIGCHI2020 submitted for peer review.
• Computer Vision Engineer - Insylo Tech SLU (contract): March 2018 - June 2018
  • Developed Computer Vision Pipelines for Volumetric Estimation of Silos using 2D and 3D images.
  • CV Techniques such as Depth from Focus and depth from Laser Mesh Projection were implemented.
  • Researched the performance of GANs for generation of depth maps using 2D (monocular) images.
  • Assessing the feasibility of depth sensing cameras (Astra Pro, Intel Realsense) in a production setting using Android device and Raspberry Pi.
• Software Engineer in Machine Learning: Facebook: September 2016 - December 2017
  • Worked with Connectivity Labs and mentored by Prof. Ramesh Raskar.
  • Researched and implemented a pipeline for Visualization of Learned Features of a CNN based on SGD to improve model training for SegNet, VGGBn and UResNet architectures.
  • Created Data Annotation tools using Qt5 used by a team (5) of GIS Analysts.
  • Optimized rendering of 2D vector geometries on an Android App. This optimization made it possible for the Android Application to be used by low end phones and reduce costs by a factor of 10.
  • All the codes are open sourced to facebookresearch/street-address
• Software Engineer Intern: iSENSES INC. January 2016 - January 2017
  • Developed a Machine Learning pipeline for Disguised Face Detection.
  • Implemented a SegNet based feature detector to identify facial action units which were then used to classify disguised faces using an SVM Classifier.
  • Entire pipeline was optimized and implemented on a an FPGA and materialized into a product.

Teaching Experience

• Neural Networks and Deep Learning - CSCI 5922: Jan 2019 - May 2019
  • Supervisor: Dr. Nicholas Dronen
  • Here I designed and implemented pedagogical versions of machine learning architectures from scratch in python.
  • Algorithms such as Auto Differentiation, Convolutional Neural Networks and Recurrent Neural Nets are implemented from ground up using numpy and numba in python allowing students to learn how mature deep learning frameworks works underneath.
• Introduction to Computer Programming with C and Python - CSCI 1300: Sep 2018 - Dec 2018
  • Supervisor: Prof. Elisabeth Stade
  • This course teaches high school level math topics like statistics, geometry, probability, and calculus using python.
  • My responsibilities include creating and grading assignments and solving doubts. See less

Publications

Robocodes: Towards Generative Street Addresses from Satellite Imagery.

Ilke Demir, Forest Hughes, Aman Raj, Kleovoulos Tsourides, Divyaa Ravichandran, Suryanarayana Murthy, Kaunil Dhruv, Sanyam Garg, Jatin Malhotra, Barrett Doo, Grace Kermani, Ramesh Raskar. "Robocodes: Towards Generative Street Addresses from Satellite Imagery". CVPR, IEEE 2017.

Generative Street Addresses from Satellite Imagery.

Ilke Demir, Forest Hughes, Aman Raj, Kaunil Dhruv, Suryanarayana Murthy Muddla, Sanyam Garg, Barrett Doo, Ramesh Raskar. "Generative Street Addresses from Satellite Imagery". ISPRS 2018.

A Holistic Framework for Addressing the World using Machine Learning.

Ilke Demir, Forest Hughes, Aman Raj, Kaunil Dhruv, Suryanarayana Murthy Muddla, Sanyam Garg, Barrett Doo, Ramesh Raskar. "A holistic Framework for Addressing the World using Machine Learning". CVPR 2018 Workshops.

Awards

• Best Paper Award for the research paper “Robocodes: Towards Generative Street Addresses from Satellite Imagery” in CVPR 2017 workshop on Earthvision.
• MS Imagine Cup 2017 Korea Semi-Finalist for our project: COBRIX