Backend systems, data pipelines, and AI-enabled infrastructure built for real engineering workflows.

Key Details

• Built data pipelines to process synchronized audio and visual signals so the model could reason across both modalities at once.
• Developed custom algorithms to detect hand positions and fret interactions, translating guitar movement into structured information.
• Designed the system as a full pipeline rather than a narrow model demo, connecting detection, feature extraction, and tab generation into one workflow.

Key Details

• At the supplier level, the system classifies fruits into perfect and imperfect categories so imperfect produce can still be sold at lower prices instead of being discarded.
• At the consumer level, the model detects fruits that are approaching expiry, making it easier to act before food turns into waste.
• The concept is built around practical value: not just identifying defects, but extending usability, preserving inventory value, and supporting more responsible consumption patterns.

Future Plans

• Develop a mobile application so classification and expiry detection can be used in more accessible real-world scenarios.
• Expand the food dataset to improve robustness across more fruit types, conditions, and edge cases.
• Evolve the system toward full lifecycle food waste monitoring, from supplier handling to consumer usage.

Key Details

• Developed backend services that process contest submissions and compute final rankings efficiently.
• Built a React interface that makes result visualization and participant standing review straightforward.
• Designed scalable evaluation logic so the platform remains practical even as the number of participants grows.

Key Details

• Built backend services using Spring Boot to manage application logic and project state.
• Implemented real-time communication through WebSockets so room changes could be reflected collaboratively.
• Containerized the application using Docker to keep development and runtime environments consistent.