PhD candidate at National Taipei University of Technology working on multi-agent deep reinforcement learning for UAV-assisted networks, reconfigurable intelligent surfaces (RIS), and space-air-ground integrated networks (SAGIN). 🏆 4.00 / 4.00 GPA
Specialty domains: OFDM PHY · MIMO · LDPC · 5G NR · RIS · Federated Learning · MADDPG · PPO · O-RAN
GPU-accelerated MMSE channel estimation for OFDM-based 6G PHY. NumPy + CuPy backends, precomputed MMSE weights mapped to Tensor Core-friendly complex64 GEMM, designed as a building block for NVIDIA Aerial cuPHY pipelines.
Result: ~9 dB MMSE gain over LS, matches theoretical 10·log₁₀(N/L) bound · 7/7 unit tests passing · reproducible benchmarks
Deep RL for 5G NR link adaptation. Self-contained PPO in PyTorch, OLLA industry baseline, 28-index MCS table from 3GPP TS 38.214, non-stationary SNR with mobility/handover scenarios. Sionna integration path documented.
Result: PPO learns competitive policy from scratch with ~3 min CPU training, fair head-to-head vs OLLA · 15/15 unit tests passing
Federated learning for CSI feedback compression. CsiNet autoencoder + FedAvg under non-IID channel statistics, aligned with 3GPP Release 18 AI-RAN study item.
Result: FedAvg matches centralised performance (~−2 dB NMSE) and beats local-only by ~2 dB · 16/16 unit tests passing
ris-beamforming-optimizer— RIS phase optimization, manifold + deep learning algorithmsoran-resource-allocation-xapp— O-RAN xApp-style resource scheduling with DRL
8+ IEEE publications in AI-native wireless networks, covering:
- Hybrid federated learning with MADDPG for UAV-assisted access networks
- Reconfigurable intelligent surface optimization for 6G
- SAGIN architectures with Starlink LEO integration
- Channel estimation and beamforming for next-gen PHY
🔗 [Google Scholar] · [ORCID] · [ResearchGate]
- Research internships in AI/ML for wireless at NVIDIA, MediaTek, Qualcomm, Foxconn — summer / full-year
- R&D collaborations on AI-native PHY, federated learning for RAN, multi-agent DRL for networks
- Quantitative engineering roles — building a production algorithmic trading system on QuantConnect since 2024
- 🎓 PhD candidate, National Taipei University of Technology — 4.00 / 4.00 GPA
- 🟢 NVIDIA NGC 6G Developer Program — Member, 2026 cohort
- 👨🏫 Advisors: Prof. Hsin-Piao Lin · Assoc. Prof. Rong-Terng Juang
📧 Email 💼 LinkedIn 🎓 Google Scholar 📍 Taipei, Taiwan · 🇫🇷 🇬🇧 🇹🇼