Shashwat Pandey

Master Thesis @ SICK AG | Edge AI for Robotics | Computer Vision & Autonomous Perception Professional

About Me

I am a Master's engineer in Perception, Computer Vision, and Edge AI for Robotics with over three years of R&D experience. My focus is translating complex vision models—such as 6D Pose Estimation and Semantic Segmentation—into reliable, real-time industrial solutions. I manage the full deployment lifecycle: from system development using ROS/ROS2, C++, and Python, to model optimization via TensorRT, ONNX, and Deepstream SDK. Currently, I am developing safe human-robot solutions at SICK Sensor Intelligence, leveraging embedded AI on 3D Time-of-Flight (ToF) cameras. I excel at integrating high-performance AI on constrained edge hardware.

Technical Expertise

Core Robotics & OS

ROS2 Python C++ Ubuntu Docker
  • 95%
  • 90%
  • 95%

AI & Deployment

PyTorch TensorFlow Python OpenCV ONNX Runtime
  • 90%
  • 85%
  • 80%

Perception & Vision

OpenCV Open3D ROS NumPy
  • 90%
  • 85%
  • 95%

Edge AI & Hardware

TensorRT ONNX Runtime Raspberry Pi Arduino NVIDIA Jetson
  • 85%
  • 80%
  • 75%

Sensors & Data

Intel RealSense SICK Sensors OpenCV Open3D
  • 85%
  • 80%
  • 85%

Software Architecture

Docker Ubuntu Git GitHub
  • 85%
  • 80%
  • 90%

Relevant Experience

Master Thesis Student (Safe Robotics / Edge AI)

Waldkirch, Germany | Sep 2025 - Present

SICK Sensor Intelligence

  • Developing an Edge AI perception system for human detection using 3D Time-of-Flight (ToF) camera data (SICK safe Visionary2), prioritizing industrial safety and accuracy.
  • Focused on model optimization and quantization using ONNXRuntime for real-time inference on embedded AI accelerators directly on camera hardware.
  • Integrating real-time object detection, segmentation, and 6D pose estimation into a production pipeline.

Computer Vision Intern (Full-time)

Ditzingen, Germany | Mar 2025 - Aug 2025

TRUMPF GmbH + Co. KG

  • Implemented a robust industrial visual inspection pipeline using YOLOv8 and SegFormer-B3, achieving significant gains in segmentation accuracy for complex manufacturing geometries.
  • Engineered a low-latency real-time system on Raspberry Pi, integrating camera capture and inference for quality inspection on constrained edge hardware.
  • Designed advanced transfer learning strategies by fine-tuning on proprietary automotive datasets.

Werkstudent – Software Developer

Augsburg, Germany | Apr 2024 – Oct 2024

KELO Robotics GmbH

  • Migrated the entire C++ robotic software stack from ROS1 to ROS2 Rolling, including creating new packages and setting up **Gazebo simulation**.
  • Led hardware setup and software integration for a Collaborative Mobile Robot, including **Python ARM API** integration.
  • Optimized system performance by debugging memory consumption and lag issues, implementing **multi-threading** and analyzing node spinning behavior.
  • Developed and maintained Docker containers for seamless testing across different ROS2 distribution versions.

Computer Vision Engineer (Full-time)

Noida, India | Apr 2022 - Jan 2023

Kritikal Solutions

  • Developed end-to-end 360° bird-eye view software stack as part of ADAS (calibration, stitching, blending).
  • Explored Apple's TrueDepth cameras; developed multi-view 3D reconstruction using photogrammetry (SfM), ICP, and point-cloud registration.
  • Deployed Deep Learning models on Nvidia TRITON server using Deepstream SDK for high-performance, real-time inference.

Computer Vision and Robotics Developer (Part-time)

London, United Kingdom | Sep 2020 - Jul 2021

Swift Robotics

  • Designed and implemented a ROS-based UV disinfection robot using Intel RealSense D455 depth cameras for autonomous room coverage and monitoring.
  • Worked on 3D mapping, localization, and obstacle avoidance modules leveraging depth maps via Stereo-SGBM and WLS filtering.
  • Established a secure wireless control system connecting Raspberry Pi nodes to a mobile app for real-time monitoring.

Key Projects

Person Following Mobile Robot (ROS2)

Sep 2023 - Jan 2024

A complete ROS2-based system for person following and obstacle avoidance. Utilized Intel RealSense D455 for sensor fusion (RGB+Depth), pose estimation, and safety maintenance.

Github Repository

AI-based Anomaly Detection (Audio)

Dec 2023 - Jan 2024

Developed an anomaly detection system using autoencoders on audio features. Achieved 95.8% accuracy via reconstruction error, ideal for predictive maintenance.

Github Repository

Winner Project: IBM Hack Challenge

Aug 2021 - Oct 2021

Developed a web app with an intelligent recommendation engine for OTT platforms, integrating real-time genre analytics and a chatbot with IBM Watson Assistant.

View Demo

Research & Education

DTLMV2 - Deep Transfer Learning Mask Classifier (Elsevier)

Published in ELSEVIER - Applied Soft Computing Journal (2022)

MobileNet V2 model trained using Imagenet weights on a custom dataset, achieving 97% accuracy. Showcases expertise in transfer learning and model validation for real-time systems.

[Full Text/Abstract Link]

Average model of DC-DC converters

Research Paper (2022)

State space analysis of practical DC-DC converters to determine the high gain operating point, considering coil resistance. (Electrical Engineering Background)

[Full Text/Abstract Link]

Education & Certifications

M.Sc. Autonomous Systems

Hochschule Bonn-Rhein-Sieg (H-BRS) | Mar 2023 - Present

B.Tech. Electrical & Electronics Engineering

Bharati Vidyapeeth's College of Engineering | 2018-2022

Key Certifications:

ML with Python Self-Driving Cars TensorFlow/Deep Learning Generative AI