Email: ac6ag0@r.postjobfree.com Phone: 734-***-****

LinkedIn: https://in.linkedin.com/in/divyanshpal

GitHub: https://github.com/divyanshpal

**** ******* ****, *** **, Ann Arbor, MI - 48105

EDUCATION -

Dual Master of Science- Robotics, E.C.E – Computer Vision Sep 2016 - Apr 2018 University of Michigan, Ann Arbor CGPA- 3.80/4

Relevant Courses: Computer Vision, Advanced Computer Vision, Machine Learning, Self-Driving Cars: Perception and Control, Mining Large Scale Graph Data, Computational Data Science, Robotics System Lab, Mobile Robotics: Methods and Algorithms. Bachelor of Engineering- Instrumentation and Control Engineering Aug 2010 - May 2014 Netaji Subhas Institute of Technology, Dwarka, New Delhi (University of Delhi) 75.6%- First Class with Distinction WORK EXPERIENCE/INTERNSHIP –

Research Associate – Bipedal Robotics Laboratory, University of Michigan May 2017 – Oct 2017

Worked on implementation of Fast Robot Optimization and Simulation Toolkit (FROST) package in C++ to reduce the time required for calculating the optimized gait of bipedal robot in real time. Systems Design Engineer - Team Indus, Axiom Research Labs Pvt. Ltd., Bangalore Mar 2016- Jun 2016

Worked in Avionics department of lunar lander and rover for GLXP competition on solar panel layout and battery configuration.

Conducted trade off studies for rover operation on the lunar surface and created test procedures. Manager - Reliance India Limited, Petrochemical Division, Jamnagar, Gujarat Jul 2014-Mar 2016

Worked as an Instrumentation Department Engineer in commissioning of Gasification plant at Jamnagar.

Resolved reliability issues, conducted Factory Acceptance Tests, reviewed engineering documentation. PROJECT EXPERIENCE –

Video Based Question Answering – Implemented a pipeline to encode content of video in a graph and then query it for answers in Python.

Each video frame is converted to a scene graph and then aggregated to form a graph that encodes the whole content of video.

Aggregation is based on two graph similarity measures – Spectral score and maximum common subgraph.

Key frame extraction was done successfully based on these graph similarity measures. Image Inpainting using GAN- Implemented an architecture for completion of an image based on Generative Adversarial Networks.

Implemented an encoder-decoder based architecture, which had a local and a global discriminator to produce both locally continuous and globally consistent images in Tensor Flow on CelebA dataset.

Implemented various GAN improvement techniques based on the paper ‘Improved Techniques for Training GAN’. Localization and Counting of Cars in a single snapshot- Implemented and trained network to estimate number of vehicles and estimate 3D bounding box in a scene given a snapshot (synthetic image, LiDAR data).

Implemented YOLO (You Only Look Once) network and trained it from scratch to estimate number of vehicles in 21 classes of data. Achieved mean absolute error value of 0.9124.

Implemented Multi View 3D Object Detection Network for Autonomous Driving for estimating 3D bounding box on each vehicle. Error Analysis of Sematic Segmentation using FCN- Performed a thorough and in depth quantitative and qualitative analysis on computer vision task of semantic segmentation using Fully Convolutional Networks (FCN).

The network was implemented in TensorFlow and resulted in a mean intersection over union (IoU) of 66.9% on PASCAL-VOC 2012. Simultaneous Localization and Mapping - Implemented SLAM and A* algorithm to autonomously navigate through a maze in C++.

Used LIDAR for environment mapping and implemented action model, sensor model and particle filter for localization.

Implemented A* algorithm to explore and navigate the robot within the maze. Graph SLAM in Dynamic Environment- Implemented Smoothening and Mapping SLAM algorithm on Victoria Park Dataset.

Implemented a robust graph simultaneous localization and mapping algorithm to a dynamic environment with moving landmarks. Gripper Design and Arm Control - Controlled a six DOF arm for picking and placing objects in its vicinity.

Implemented Forward Kinematics and Inverse Kinematics using DH Convention in Python.

Used overhead hanging camera and Open CV for blob detection and finding their location with respect to arm. Design and Control of Balance Bot - Used Beagle Bone as controller and PID control to create a balance bot and control its yaw, pitch, roll angle and velocity. Designed complete mechanical and electrical layout of the balance bot.

Used Lightweight Communication and Marshalling (LCM) system to transmit data to a server where all the data was plotted. Adaptive Cruise Control- Implemented adaptive cruise control on a simple model of car using SIMULINK, Stateflow and S-functionality.

Controller designed for 3 modes- manual mode, position control (when following a vehicle) mode and velocity mode.

Cars broadcast their positions and receive information of other cars on CAN network bus. SKILLS –

Programming Languages: C++, C, Python, MATLAB Tools: Linux command line, GIT Software: Tensor Flow, Hadoop, SIMULINK, LabVIEW, Autodesk Inventor Languages: Hindi, English, German (B1 level) DIVYANSH PAL

Contact this candidate