Taipei–Keelung Metropolitan area
Software Engineer and Applied AI Researcher with a Ph.D. from USC, specializing in advanced system architectures and mathematical modeling. Brings 4+ years of high-impact industry experience at MediaTek, driving innovations in 6G communications that led to granted patents and key 3GPP standard contributions. Most recently, I spearheaded context engineering strategies, optimizing LLMs to process massive unstructured datasets. Passionate about combining rigorous theoretical analysis with scalable GenAI deployment to build robust, high-performance infrastructure and data-centric applications.
- Designed a schema-driven context engineering architecture for a massive-scale RAG framework, optimizing API inference efficiency while establishing a verifiable source tracking mechanism to meet strict Med-Legal compliance. - Architected and open-sourced the Superinsight AI Benchmark, a standardized evaluation framework for frontier LLMs in high-stakes Med-Legal domains, enhancing model selection transparency and driving enterprise-level adoption.
- Spearheaded advanced research initiatives in GenAI/LLM/Computing and Sustainability, exploring their implications for next-generation systems (e.g., 6G, AI-RAN). - Developed a Python-based performance evaluation platform for analyzing the efficiency and effectiveness of LLM use cases within an AI-RAN system. - Pioneered the design of a sustainable 6G communication and computing system architecture, substantiated by granted patents, technical report and academic publications. - Contributed to 6G standardization through 3GPP delegation, proposing and advocating for innovative use cases such as LLM-based AI Agents and Sustainable Systems.
Projects on AI, federated learning, distributed computing, information theory: - Efficient Secure Aggregation for Federated Learning: Reduced federated learning overhead with LightSecAgg, preserving privacy and dropout resilience through one-shot reconstruction. - Online Learning for Unknown Edge Computing Networks: Formulated a contextual multi-armed bandit for optimal offloading in unknown edge computing environments and developed an asymptotically optimal online learning policy. - Security for Large-Scale Distributed Systems: Proposed novel coding schemes for secure computation in distributed systems with potential faulty workers and designed PolyShard, an iterative coded storage/computation scheme for enhanced blockchain security. - Optimal Load Allocation for Timely Coded Edge Computing: Modeled variability in edge cloud speeds and developed an optimal dynamic computation strategy for timely coded execution. - Optimal Task Scheduling for Large-Scale Heterogeneous Edge Computing: Proposed a novel virtue queueing network and an optimal communication-aware scheduling policy for online task scheduling in heterogeneous edge networks with graph-based workloads.
Self-Sustainable OFDM Transmissions with Smooth Energy Delivery: - Investigated and implemented the frame-theoretic operation for reducing PAPR of OFDM transmission - Designed a novel architecture for self-sustainable OFDM transmission with lower PAPR in the CP while maintaining nearly the same self-sustainability of prior self-sustainable OFDM techniques - Featured in the IEEE Signal Processing Magazine (https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7814327)