Suryansh SHARMA

Experience

• 

  University of Technology Sydney

  Nov 2020 - Jul 2024

  •Worked extensively with Lidar as the primary perception source. Worked on ground plane removal-based algorithm that works on point density to segment ground and non-ground points, downsampled points for computational efficiency, and integrated with the robot navigation stack to create a costmap. This segmentation method help maintain the desired publish rate on a higher resolution point cloud.Collaborated on the Adaptive Trajectory Planner, also creating a version that supports four-wheel steering trajectories to perform spot turns for goal positions situated behind the robot. With the help of inflation costs and the right trajectory definition (and proper cost functions), the robot can drive autonomously through different terrains. Recently I made an additional node that would provide reverse velocity commands when there is not enough momentum to perform a spot turn integrated with the twist-mux package.• Managed systems running on Ubuntu with ROS, deploying services using containers to avoid dependency issues.•Developed multi-threading packages for handling sensor data and other concurrent processes using techniques like bind boost and boost thread to run functions simultaneously within a node.• C++, Python, and Bash: Utilized C++ for high-performance tasks such as local planning and joystick interface, and Python for tasks like GUI development and twist-mux locks. Transitioned from bash to docker-compose for package deployment ease.•Created an NLP model to convert natural language commands into signal temporal logic formulas for initiating robot path planning. This model achieved higher accuracy than previous models in the field and was the basis for a paper published in ACRA 2023. Show less Software Engineer-AI(Deep Learning) at School of Computer Science. The project utilizes Deep Learning techniques(object detection) to perform Search And Rescue Operations (SARS) in flooded areas; detect objects of interest and rescue them by dropping life-saving pods. The project was in collaboration with Australia4Innovation and LQTDU Vietnam.Developed computer vision model tailored to different terrains (forests, water bodies, hilly areas) to minimize false positives and enhance environment-specific detection for detecting people drowning in floods.Winner of NSW State iAwards 2024 in goverment and public sector. Show less

  • Software Engineer

    Aug 2022 - Jul 2024

  • Software Engineer

    Nov 2021 - Aug 2022

  • Casual Academic Tutor- Data Processing Using Python

    Jul 2021 - Oct 2021

  • Casual Academic Tutor- Deep Learning

    Feb 2021 - Jun 2021

  • Research Intern

    Nov 2020 - Feb 2021
• 

  DroneShield

  Sept 2024 - now

  Artificial Intelligence Software Engineer

  AI Software Engineer in the Radio Frequency AI Team at Droneshield.- I am currently working on tracking and detection algorithms.