Poh Wencuan

Led a cross-functional team including data scientists, data engineers, analysts, and software engineers to develop and deploy end-to-end manufacturing AI solutions worldwide. Focused on Smart Use Case Methodology and CRISP-DM for machine learning projects. Managed sprint planning, feature prioritization, and retrospective meetings, improving the software development process. Optimized algorithm performance and designed machine learning architecture for real-time model monitoring using Google Cloud Platform and MLOps best practices. Mentored R&D teams on MLOps, software development, and deployment. Achievements include stabilizing AI/ML prediction pipelines with 99% uptime and 1-minute latency handling 6 million sensor predictions per hour, designing a custom NLP clustering model to identify root causes of pipeline failure, and collaborating to improve anomaly detection and metrology models.

Led a cross-functional team to translate machine learning models into end-to-end manufacturing solutions deployed worldwide using Smart Use Case Methodology and CRISP-DM. Conducted sprint planning, managed feature priorities, and facilitated continuous software process improvement. Optimized algorithm performance and designed machine learning architecture pipelines with real-time model lifecycle monitoring using Google Cloud Platform and MLOps best practices. Mentored R&D teams and collaborated with vendors to enhance industry 4.0 solutions. Achieved 99% uptime with 1-minute latency for sensor predictions processing 6 million sensors per hour and designed custom NLP clustering methods to improve dataflow failure analysis.

Translated key requirements from SOC engineers into AI solutions and led a team to develop a pre-silicon validation modeling platform with REST API and DAG programming. Developed an AI model for third-party SOC IP signal matching achieving 85% automation with 100% true positive rate. Delivered an expert model architecture using NLP and generic database storage to convert human queries into SQL commands. Improved validation speed significantly and supported SOC engineers' productivity.