Cass Sheng

• Architected and developed the backend infrastructure using Go and Google Cloud Platform (GCP) to support scalable AI-driven applications.
• Designed and built a desktop AI assistant app using C++, LLAMA, Python, Docker, and TypeScript, enabling seamless interaction with AI models.
• Developed and deployed a React-based website, ensuring an intuitive user experience and seamless integration with backend services.
• Led full-cycle product development, from prototyping to production, leveraging expertise in AI, cloud infrastructure, and modern web technologies.
• Deployed an AI assistant for cancer patients, optimizing healthcare interactions and providing support through advanced NLP and ML models.
• Implemented CI/CD pipelines and containerized microservices for efficient and automated deployments across cloud environments.
• Collaborated with cross-functional teams, including AI researchers, designers, and product managers, to align technical solutions with business goals.
• Drove technical decision-making and established best practices in system architecture, security, and performance optimization.

• Architected and developed server-side applications, databases, caches, algorithms, networking components, and rendering pipelines, ensuring robust and scalable solutions.
• Designed and implemented an efficient rendering software for displaying real-time 3D point cloud data on mobile devices using OpenGL, enhancing the visual representation and user experience.
• Designed and implemented APIs for integration with various services and systems, enhancing interoperability and functionality.
• Created and implemented a robust network architecture utilizing AMQP and WebSockets for seamless communication between tracking modules and client interfaces.

• Spearheaded the design and development of first-generation network components from scratch, securing a 1st place win at the FRST competition.
• Architected a scalable backend using Python FastAPI, with enhancements like Redis caching and Docker containerization, deployed with Kubernetes.
• Implemented a centralized admin panel and utilized GCP performance logging for comprehensive system monitoring.

• Developed a predictive model to determine internal layouts of single-family residential houses using external images.
• Implemented a feature extraction model using object detection to parse external house images and used to obtain layout predictions.
• Deployed the model on GCP, serving predictions via a Flask server connected to a mobile app, facilitating real-time access and usability.