Matter Motor

---

Matter Motor is a premium electric motorcycle start-up based in India, with a focus on emerging markets. I was an intern from 2022-2024 It employs over 600 employees. Matter Motor launched it's first motorcycle Matter Aera. This is the first electric motorcycle with liquid water cooling and gears.

Matter is consolidated into 5 departments:

1. Matter Energy
2. Matter Entel
3. Matter Link
4. Matter X
5. Matter Drive

For more information go to https://www.matter.in/

This study investigates the laser beam spot welding of copper to Hilumin® (nickel-plated steel) for electric vehicle (EV) battery packs. Due to copper's high thermal and electrical conductivity, laser beam welding (LBW) is preferred over traditional methods like TIG and plasma welding, which struggle with copper’s heat dissipation and reflective properties. Despite LBW's advantages, challenges such as high reflectivity and thermal conductivity of copper hinder joint efficiency, leading to issues like energy wastage and porosity. This research aims to optimize LBW parameters for copper-Hilumin® joints, focusing on shear strength, pull strength, and visual weld quality to enhance structural integrity and energy efficiency in EV battery packs. Our investigation showed a ~72% in joint efficiency.

The details of the parameters are confidential since the international patent is currently under the filing process.

The application of these patents will be for copper busbar to Hilumin® cased lithium ion cell to meet electric vehicle regulation. A higher joint efficiency will ensure longevity and can survive against vibrations in the vehicle. Another advantage is lower resistance and higher power efficiency. This work was done under the mentorship of Dr. Raghavendra Darji.

For this patent I delved into great depth in chemistry and composite materials properties to help create a composite foam that can insulates battery pack from ambient heat, while also safeguarding it from vibrations of the electric motorcycle. I had to put on a process orientated approach with a focus not only on performance, but also ease of manufacturing. For this project, I worked under the mentorship of Divij Vaishnav and Dr. Raghavendra Darji.

The material comprises two compounds. They include Poly-propylene glycol, and modified poly-iso-cyanate with one other undisclosed material.

The details for the material are still under a non-disclosure agreement, expected filing in 6 months.

 On Matter Motor’s production line, I noticed circuit boards piling up at the quality inspection area, creating a bottleneck despite rapid welding apparatus. To address this, I proposed automating quality inspection with a camera to identify faulty welds.

I set out to build a high-accuracy, single-shot detector, exploring architectures like YOLO, ResNet50, DPM, and R-CNN. After selecting YOLOv8 and fine-tuning its hyperparameters, I encountered a surprising issue—high reflectance from the copper base metal caused noisy images. Viewing it as a materials challenge, I found that copper reflectance was lowest at a wavelength of 400 nm and added a blue LED light ring around the camera to reduce noise. With this adjustment, the model achieved over 90% accuracy.

The system cut inspection time by 60%, allowing quality assurance workers to focus on other tasks while maintaining ethical workforce considerations. Although the model’s accuracy fell 5% short of manual inspection, the experience taught me to rethink bottlenecks and innovate with a systems-thinking approach.

 I kickstarted the development of a custom object detection algorithm for Matter's self-driving division, laying the groundwork for its autonomous mobility solutions. Using the YOLOv8 framework, I designed and optimized the model to ensure real-time performance, fine-tuning hyperparameters to strike a balance between high accuracy and computational efficiency.

A key aspect of the project was adapting the model for edge computing. I integrated it with the on board Snapdragon 700 SoC, leveraging edge computing principles to address hardware constraints. This involved reducing the model’s computational cost and optimizing its latency to enable the local processing of visual data. By eliminating the reliance on cloud connectivity, the system achieved lower response times and enhanced energy and computational efficiency.