Abhiraj Gupta

Working on Software supporting Automated Vehicles and terminal planning

As a Software Engineering Intern at PSA International, I played a key role in automating and optimizing the AGV (Automated Guided Vehicle) Routing Test Framework. By developing an automated testing system, I expanded test coverage from 600 to 4,500 scenarios while reducing execution time by over 99%, significantly enhancing the efficiency and accuracy of AGV routing validation. To further improve system reliability, I integrated automated traffic rule validation modules, ensuring consistent and error-free routing behavior checks while minimizing manual intervention. Additionally, I developed dynamic test case automation scripts, reducing manual effort by 90% and streamlining the verification process for 4,500+ complex scenarios. Beyond automation, I redesigned data parsing and configuration structures, enabling the system to process 1,000+ complex map data points with 60% faster processing time. These improvements contributed to a more scalable and efficient system for AGV routing and logistics operations. Through this experience, I honed my skills in automation frameworks, large-scale testing, data processing, and system optimization, while contributing to PSA’s mission of advancing efficiency in automated port logistics.

As an Undergraduate Research Student, I explored Neural Architecture Search (NAS) for edge computing, focusing on optimizing deep neural networks (DNNs) to enhance efficiency on resource-constrained devices. My research involved identifying and fine-tuning network architectures tailored to specific hardware and application requirements, bridging the gap between hardware limitations and software performance. I implemented NAS techniques in real-world deep learning models, evaluating their efficiency through experimental data collection and practical demonstrations. A key aspect of my work was refining algorithmic efficiency, with a focus on "once-for-all" algorithms, which enable adaptable model architectures for varying deployment constraints. This research deepened my understanding of hardware-aware AI, model optimization, and computational efficiency, equipping me with hands-on experience in machine learning, NAS frameworks, and edge computing.

During my internship at PSA Singapore, I collaborated with the AI team to enhance operational efficiency in the port logistics sector. My primary focus was on identifying and mitigating bottlenecks in quay crane operations, particularly those related to the secondary trolley's idle times. By analyzing real-time log data from Tuas port, I identified correlations between the idle times of quay cranes and the locations of Automated Guided Vehicles (AGVs). This analysis led to the development of re-routing strategies that minimized delays and optimized workflow. In addition to operational analysis, I partnered with the Proof of Concept (POC) team to streamline the stowage marking process within barges. I developed an advanced logic system for optimal container planning across vessels and created a user-friendly Graphical User Interface (GUI) to enhance the understanding of stowage arrangements. This interface facilitated efficient container placement, contributing to smoother and more effective operations. Key contributions: Improved quay crane efficiency by identifying and addressing idle times related to AGV positions. Developed re-routing strategies to reduce delays and optimize port operations. Created a logic system and GUI for better container planning and stowage within barges. This internship provided me with valuable insights into the practical applications of AI and algorithmic solutions in real-world scenarios, further solidifying my interest in leveraging technology to solve complex logistical challenges.