Shubham Parab

· Developed an innovative approach for 3D underwater scene reconstruction by integrating 3D Gaussian Splatting (3DGS) with a lightweight neural network to accurately model underwater light effects.
· Reduced training time by over 90% compared to previous methods, enabling rapid model deployment.
· Increased rendering speed to real-time performance (up to 241 FPS), making the solution ideal for aut

· Reduced pole-verification time from months of manual review to just eight minutes per 100 poles through automated suitability assessment.
· Reduced manual planner workload by 97.5% by developing an end-to-end ML pipeline and Streamlit dashboard for automated LoRaWAN gateway site surveys.
· Achieved 98% accuracy and 99.2% precision across 3,070 validation images using SAM3-based object detection/

• Automated the large-scale generation of 3D synthetic data using FLAME, generating 10,000+ annotated face meshes and drastically reducing manual dataset creation effort.
• Executed mesh-level geometry processing with Python and NumPy, streamlining data handling.
• Applied preprocessing techniques, including scaling, normalisation, and mesh simplification, to prepare data for model training
• Eval

• Engineered end-to-end Computer Vision and Robotic Perception modules using Python and OpenCV for Retail Automation and Industrial Robotics.
• Developed and deployed Object Detection and Anomaly Detection pipelines, enabling autonomous navigation and decision-making for omnidirectional robots.
• Led the integration of Vision-ML models with ROS/ROS2 middleware and control systems, ensuring low-lat

• Architected scalable backend modules using Python and the Django Framework, ensuring high availability for web applications.
• Implemented RESTful APIs and conducted systematic debugging to resolve legacy software issues, improving code maintainability and performance.

• Assisted in the design and layout of PCBs for embedded systems applications.
• Collaborated with the hardware team to test and troubleshoot electronic circuit designs.

• Designed and simulated electronic circuits for Embedded Systems and affordable robotics platforms using Tinkercad.
• Authored technical documentation and blog posts on Circuit Design, translating complex engineering concepts for a broader technical audience.

Coursework:
1. Data Visualisation
2. Statistics
3. Image Processing
4. Machine Learning
5. Data Science
6. Advance Artificial Intelligence
7. Deeplearning
8. Engineering For AI
9. Disertation

Sep 2025

Sep 2025

Jan 2025

Dec 2025

Final-year project: Developed a computer-vision-guided fruit-picking robot, integrating computer vision, image processing and robotic systems for automated fruit detection and manipulation.

Sep 2021

Sep 2021

Mar 2021