Anubhav Singh

Block B, Sector 53 · Noida, Uttar Pradesh, India · singh7.anubhav@gmail.com


I'm an AI and robotics enthusiast who is passionate above solving practical problems which affect humans on day-to-day basis using different AI techniques. My area of Interest includes robotic motion planning, computer vision, reinforcement learning, kinematics, navigation, etc.

One of my desires is to build an autonomous biped walking robot using reinforcement learning and computer vision.




“You don’t have to be perfect to be amazing.”

Experience

Research and Development Engineer

Orangewood Labs

It’s a design-led and a Y Combinator backed tech startup based out of San Francisco that creates customized furniture using robotics.

  • Developed a 6 DOF collaborative robotic arm with a payload capacity of 5 kg and a reachability of 860 mm using ROS framework. Implemented the kinematics and trajectory planning of the robotic arm, both in cartesian and joint space.
  • Developed a machine learning based system to compensate for the cumulative error in any motion transmission system. This invention has been patent approved with the U.S. Patent and Trademark Office (USPTO).
  • Developed a teaching pendant for the robotic arm using C++ Qt framework.
  • Worked on the 3-DoF and 6-DoF pose estimation open source ROS packages like DOPE, darknet_ros, Linemod,etc.
  • Developed an app for controlling a mobile robot using ESP32 and BLE modules.
  • Was part of the team which developed a custom version of open source bCNC software.
    My contribution was integration of the software with the google cloud platform for data(CNC design files) storage and showing it on the main page of the application.
  • Developed a speech-to-command api for controlling the robotic arm using BERT language model and google cloud platform.

Oct 2018 - Dec 2021

AI Engineer Intern

Lumiq.ai

  • Worked on tikali Energy Usage Profiling Project.
  • Built a good/bad energy usage profiling plots classifier using svm, which is successfully running in production and saving a lot of time of the data engineering team by automatically detecting the bad plots in real time.
  • Built a dashboard to track the data deletion process by the clients.
  • Also, designed a Nifi flow to initiate the data deletion process on client’s request.
  • Tools Used: Machine Learning, Nifi, Graylog, Airflow, Deep Learning, OpenCPU

Feb 2018 - Aug 2018

Education

Jaypee Institute Of Information Technology, Noida
Bachelor of Technology(BTech)
Computer Science and Engineering

CGPA: 7.00

Aug 2014 - May 2018

Patna Central School

Higher Secondary School(Intermediate)
Courses: Physics, Chemistry, Mathematics, Informatics Practices, Physical Education
Activities and Societies: Basketball Team, Volleyball Team

Score: 91

Mar 2011 - Mar 2013

Patliputra Central School

Secondary School/High School
Activities and Societies: Science Exhibition, Cricket, Basketball

Score: 95

Mar 2003 - Mar 2011

Skills

Programming Languages & Tools
FRAMEWORKS & LIBRARIES
  • ROS 1(Kinetic, Melodic, Noetic)
  • ROS 2 (Foxy Fitzroy)
  • MoveIt
  • Pytorch
  • Tensorflow
  • OpenCV
  • Deep Learning (mostly RL)
  • Point Cloud Library (PCL)

Robots I worked with

PUBLICATIONS & RESEARCH

(Patent) SYSTEM AND/OR METHOD FOR ERROR COMPENSATION IN MECHANICAL TRANSMISSIONS

  • To compensate cumulative error in any gear-based motion transmission system using machine learning based mechanism.

Projects

Github contributions in the last year

anubhav's Github contribution chart

Control and simulation of a 5 DOF Robotic Arm using ROS

  • The objective of the project was to control a 5 DOF Robotic Arm using ROS Moveit Configuration package.

6 DOF Collaborative Robotic Arm

  • Project concerned with building a 6 DOF collaborative robotic arm using ROS 1 framework. The project was initially started with ROS 1 Kinetic but later on migrated to ROS 1 Noetic distribution.
  • Developed the kinematics and trajectory planning(both in Joint as well as Cartesian space) of the robotic arm using ROS MoveIt package.
  • Generated custom IKFast plugin for the robotic arm using OpenRAVE motion planning software.
  • Setup ros_control and developed a hardware interface for the robotic arm. We used CAN Bus Protocol for data transmission and communication purposes.

Object Segmentation and clustering in realtime 3D Point Clouds using PCL

  • Techniques used are Voxel filtering, Statistical outlier removal, RANSAC plane segmentation, Euclidean clustering for object segmentation, etc.
  • Objective was to locate centroid in each segmented object and published the centroids’ coordinates on a ROS publisher of type visualization_msgs/MarkerArray.

Robotic Arm Teach Pendant GUI

  • Developed a teaching pendant for a robotic arm using C++ Qt framework.
  • The features added were joint based control, trajectory planning(Joint space as well as Cartesian space), hand teaching(gravity compensation), jogging(joint, end effector and joystick), robot 3D visualization, sensor data visualization, speech to robot command, etc.

Automated Guided Vehicle(AGV)

  • The objective of the project was to build a portable robot that can be used in industrial applications to transport heavy materials around a large industrial buildings using computer vision and convolution neural networks.

Image Classifier

  • The objective of this project was to check if an energy usage profiling plot/chart is correct or is an error thrown by server, i.e. something went wrong while building the energy usage plot.

Product Recommendation System

  • The objective of the project was to recommend products to a user based on its similar neighbors.