Harman Kumar

Product Engineer

New York, New York, United States8 yrs 6 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Best Thesis Award for Autonomous Vehicle project
Expertise in Computer Vision and Localization
Strong background in Software Development and Algorithms

Stackforce AI infers this person is a skilled engineer in the Automotive industry with a focus on Computer Vision and Localization technologies.

Contact

Skills

Core Skills

Computer VisionSoftware DevelopmentLocalizationMapping

Other Skills

Graph SLAMICPKD Tree3D Point cloudIMUGPSSensor MeasurementsProduct VisionProduct ManagementProduct DesignCommunicationC++Data StructuresProgrammingArtificial Intelligence

About

During my Undergraduate studies, I explored the fields of Computer Vision, Artificial Intelligence, Computational Geometry and Data Structures in great depth. For my Undergraduate Thesis, I worked on the localization, perception and planning unit of an autonomous vehicle and won the Best Thesis Award for the academic year 2017-18

Experience

8 yrs 6 mos

Total Experience

2 yrs 10 mos

Average Tenure

5 yrs 10 mos

Current Experience

Jane street

International ETF Desk

Aug 2020 – Present · 5 yrs 10 mos · New York City Metropolitan Area

Cruise automation

Research Engineer, Localization and Mapping

Jun 2018 – Jul 2020 · 2 yrs 1 mo · San Francisco Bay Area

LocalizationMappingComputer VisionSoftware Development

Snap inc.

Research Engineer

Oct 2017 – May 2018 · 7 mos · London, United Kingdom

Computer Vision
Software Development

Computer VisionSoftware Development

Samsung electronics hq (south korea)

Research Intern

May 2016 – Jul 2016 · 2 mos · Seoul, Korea

Localization and Mapping for Autonomous Vehicle
For the construction of accurate environmental model from sensor measurements (3D Point cloud, IMU, GPS), Graph SLAM was implemented.
SLAM Frontend:
For odometry estimation, Scan Matching was done using the ICP algorithm.
Efficient loop closure detection algorithm based on KD Tree search was implemented.
Normal vector estimation, Keypoint estimation and Feature Correspondence were used for estimating similarity of point clouds to serve as initial guess for ICP.
SLAM Backend:
The iSAM algorithm was used to optimize the factor graph by converting it into a non linear least squares optimization problem.

LocalizationMappingGraph SLAMICPSensor Measurements