Swapnil Gaikwad

Skilled in mastering data, and study-related procedures and protocols to ensure service excellence. Well-versed in project management and research planning, with proven talents. Pursuing a new professional challenge and offering key abilities and experience.

• University Mentor, University of Western Australia, 02/24, Present
• Team Lead for Sponsors, Utkarsh Cultural Festival, SK Somaiya College of Arts College and Science, 04/19

Prototype Acoustic Traffic Monitoring System, Implemented on a Low-Power Edge Processor

• The primary objective was to design a low- power acoustic traffic monitoring system using the Xylo Audio Processor.

Key Goals:

• Constructing a comprehensive dataset for vehicle classification
• Implementing a Spiking Neural Network (SNN) Architecture tailored for low-bit depth vehicle classification. Successfully deploying the model to the Xylo Audio processor for real-time inference
• Assessing metrics such as power consumption, accuracy, false detection rates and miss rates

Key Achievements:

• Leveraged the DCASE2024 challenge dataset with 14,999 audio recording from urban environments, resulting in a final dataset of 500 samples per class (car, Commercial Vehicle, background)
• The Final SNN achieved an accuracy of 87.1% with an F1 Score of 86.9%, demonstrating effective vehicle classification capabilities.
• The system maintained low power consumption during operation, averaging around 224 uW during inference.
• The quantized Model was deployed on the Xylo processor without performance degradation

Tools / Skills Used:

Rockpool Library, Python, Google Collab, Librosa, PyTorch, Scikit Learn, Weights and Biases. (For experiment tracking and managing model versions).

• Global Excellence Scholarship, University of Western Australia, 07/23
• Process Pro, Teleperformance PVT LTD, 05/18

• The Complete 2023 Web Development Bootcamp | Dr. Angela Yu
• Data Structures and Algorithms | Coding Ninjas
• Beginning C++ Programming – From Beginner to Beyond | Dr. Frank Mitropoulos