Che-Cheng (Reeve) Wu

Served as a Microsoft Student Expert for DevDays Asia 2025, preparing a workshop and designing the "Next Big AI Student Zone" booth experience to showcase Microsoft's latest technical resources and support for students.

• Prepared a DevDays Asia 2025 workshop that helped 30+ developers build end-to-end AI prototypes with Azure OpenAI and Cognitive Services.
• Designed booth content for the "Next Big AI Student Zone," presenting Microsoft technology resources, learning paths, and student support programs.

Delivered generative AI solutions on Azure and Power Platform, turning emerging AI capabilities into production-ready proof-of-concept systems for enterprise scenarios.

• Delivered 40+ generative AI projects, accelerating enterprise adoption of AI-driven workflows and validating new features through PoC solutions.
• Designed RAG and agentic frameworks for domain-specific automation and integrated them into scalable production-ready solutions.
• Built an automated video-segmentation pipeline to bypass system limits and improve reliability for long-form transcription workflows.
• Developed a multi-agent system for autonomous OpenAPI invocation to reduce manual integration overhead.

Built evaluation tooling for enterprise Text-to-SQL systems and helped improve data quality analysis workflows for factory operations.

• Automated Text-to-SQL benchmarking across multiple database schemas through a scalable evaluation pipeline.
• Engineered a factory database evaluation workflow for quality monitoring and operational analysis.
• Applied Agile Scrum and Test-Driven Development to support iterative feature delivery and code reliability.

Advanced research in retrieval-augmented generation and ranking systems, focusing on hybrid retrieval pipelines, evaluation, and reliability analysis.

• Designed a hybrid retrieval pipeline using BM25, dense retrieval, ColBERT reranking, and RRF fusion.
• Achieved +35% Recall@1000 and +31% nDCG@10 over baseline BM25.
• Researched how uncertainty in LLM internal signals can be used as a retrieval criterion to improve evidence recall and response quality in RAG systems.

Built full-stack, LLM-driven urban intelligence systems combining multi-agent reasoning, geospatial analysis, and simulation workflows.

• Deployed a full-stack LLM-based multi-agent system on Azure for autonomous retrieval, visualization, and analysis of geospatial data.
• Analyzed and visualized over 30K citizen hotline records, leveraging semantic clustering to identify spatio-temporal distributions, trends, and anomalies.
• Enabled scenario-based analysis through an agent-based simulation platform integrating open mobility and urban data.

Researched LLM-based navigation and obstacle avoidance for indoor drone applications, combining sensor data with language understanding capabilities to enhance drone autonomy in complex environments.

• Designed and implemented a drone navigation system that integrates LLMs with sensor data to achieve autonomous navigation and obstacle avoidance in indoor environments.

Led technical learning experiences around cloud and AI for the student and developer community.

• Designed and delivered two workshops focused on AWS Bedrock and OpenSearch for 100+ participants.
• Received satisfaction scores of 4.4 and 4.8 across workshop sessions.
• Guided learners through foundational AI and cloud concepts as an AWS AI Practitioner course instructor, reaching 1,800+ YouTube views.

2025 AWS Educate Taiwan AI Practitioner Certification Sprint Course

youtube.com

Planned and executed career development activities, helping students and job seekers understand LinkedIn's career resources and opportunities.

Contributed to agenda planning and coordination for a student-led computing conference.

Led campus election affairs and coordinated election operations for student self-governance.

• Coordinated and supported the planning of 20+ campus events during the term.