Ansu Mathew

Email: adkmjk@r.postjobfree.com

Mobile: +971-*********

Location: Dubai

LinkedIn

Skills and areas of interests

Skills

Languages: Python, Matlab, C

Frameworks: Tensorflow, PyTorch, Keras, Scikit-Learn,Pandas, Matplotlib, Seaborn, Folium. Areas of interest

Machine learning

Computer Vision

Deep learning

Data visualization

Experience

July 2020 –till date

Projects

Enhancing machine learning skill in business intelligence through online courses (Licenses) and projects

Machine Learning

Build a model using logistic regression and decision tree to automate the loan eligibility process (real time) based on customer detail provided while filling online application form Github

Build a model to devise campaigns with better target marketing to increase the success ratio with a minimal budget Github

Data Visualization

Performed visualization to summarize the results of a survey that was conducted to gauge an audience interest in different data science topics.

Plotted a Choropleth map of the crime rate in San Francisco Research Scholar Indian Institute of Science Aug 2018 – Jun 2020 Highlights

PhD student, Video Analytics lab, IISc

PhD research area: 3D scene understanding with minimal supervision using artificial intelligence

Completed 13 credits with 9 GPA as part of coursework

Research paper on 3D Reconstruction accepted at CVPR 2020

Led a team on a project on understanding convolution neural networks for 3D tasks

Guided team members for improving 3D reconstruction by introducing novel losses

Hands on experience in implementing Principle Component Analysis, LDA, Linear and Logistic Regression, Decision Tree, Random Forest, Adaboost, GradientBoost, SVM, Clustering, neural networks, convolutional neural networks, Resnet, GAN losses. Projects &

publications

Deep learning

Understanding convolution neural networks for 3D tasks [Python, PyTorch]

Improving 3D point cloud reconstruction by introducing novel losses [Python, Tensorflow]

From image collections to point clouds with self-supervised shape and pose networks[Python,TensorFlow]

o Performed 3D reconstruction and pose estimation in a self-supervised manner.Imposed 3D geometric and pose consistency loss between point clouds for training. Utilized k nearest neighbor(knn) samples from train set as pseudo viewpoints o Publication: Our work "From Image Collections to Point Clouds with Self-supervised Shape and Pose Networks has been published in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 1132-1140(Video,Code)

3D point cloud reconstruction of objects using a single image [Python, Tensorflow] o Aims to reconstruct a 3D shape from a single image with known pose. GAN loss introduced for enforcing the constraint that multiple views need to come from the same 3D object o Explored the relationship between 2D image flipping and 3D rotation Machine learning

Gaussian mixture model for audio classification [Python] Computer vision

Motion segmentation [Python]

o Used motion to extract objects or regions of interest from a background of irrelevant details o Implemented background subtraction, temporal integration, hierarchical segmentation, K- means clustering on optical flow vectors, optical flow and Otsu thresholding

Pose estimation of the brick [Python]

o Detected the brick against any background irrespective of its orientation and dimensions in the image and estimated its pose

Assistant Professor Manipal Institute of Technology Jul 2015 – Jul 2018 Highlights

Assistant Professor in the Mechatronics department

Main subjects - Machine Vision and Image Processing, Digital Signal Processing, Digital Integrated Circuits and Control Systems

Integral member of the curriculum development team for the Mechatronics department

Took the lead in designing machine vision experiments for the newly commissioned robotics lab

Conducted workshop on 'Sherlock', a machine vision software as part of faculty development

Procured A+++ (highest rating) for excellence in teaching in 2017 Projects &

publications

Computer vision

An Investigation of Burr formation and cutting parameter optimization in Micro Drilling of Brass C-360 using Image Processing. Publication in Advances in Intelligent Systems and Computing 828, pp. 289-302 [Matlab]

o Developed an image processing technique to measure the burr size and hole circularity Internship AMD India Pvt. Ltd. Jul 2014 – May 2015 Highlights

Developed a framework for power management in battery constrained systems

Designed a Register Transfer Logic model for the Motion Estimation (ME) block of H.264

Developed a simulation model of H.264 using MATLAB to track performance uplifts

Runner up for the project submission entitled “Improving energy efficiency of Video Conferencing using Co-operative power management”, for Software Innovation Day, 2015 Projects &

publications

Computer vision

Exploiting loop perforation for improving motion estimation performance of video encoders. Publication in ARPN Journal of Engineering and Applied Sciences 10(17), pp. 7539-7544. Also, Presented at AMD Asia Technical Conference [Matlab] o Performed optimization on Matlab and Verilog

Education

The University of Texas at Austin 2020 - 2021

Post Graduate Program in Artificial Intelligence and Machine Learning: Business Applications by The McCombs School of Business

Indian Institute of Science 2018 - 2020

Robert Bosch Centre for Cyber Physical Systems, PhD Research Scholar

Course work: GPA – 9 /10 (completed 2 years of studies in pursuit of PhD in deep learning before discontinuing the program due to logistical constraints)

Subjects – Random Process, Digital Image processing, Deep Learning for Computer Vision, Neural Networks & Learning Systems and Autonomous Navigation

Cochin University of Science and Technology 2013 - 2015

Electronics, M.Tech, GPA – 9.3 /10

Secured second rank for M.Tech

Amal Jyothi College of Engineering 2008 - 2012

Electronics and Communication Engineering, B.Tec, Aggregate course percentage – 78.48%

Selected for Director’s Honors list (top 5 students in academics) consistently in every semester

Contact this candidate