Communication Systems × AI · Hsinchu, Taiwan
Chung-Yin Ho何忠穎 · “Nick” · Senior Engineer @ MediaTek
I bring AI into the heart of the modem — where cellular PHY meets machine learning, from 3GPP standards to models running on live silicon.
mtk21750@gmail.com / LinkedIn / Google Scholar / +886 989 158 362
◆ Flagship — AI-RAN Alliance @ MWC 2026
AI Transmit Diversity & Model Lifecycle Management
An AI-accelerated uplink proof-of-concept (AI TxD), exhibited as “AI for 6G Uplink” — an official AI-RAN Alliance demonstration shown at both the MediaTek and AI-RAN Alliance booths and validated with Keysight, delivering up to 30–40% cell-edge uplink gains. I owned the UE-side pipeline end-to-end — model training → embedded implementation on modem MCU/DSP → integration into a functioning smartphone — framed within a Model Lifecycle Management (LCM) architecture for AI in the air interface.
01
Summary
Senior Engineer in Communication System Design (5G-Advanced / 6G) at MediaTek, working where cellular PHY meets machine learning. My flagship work — an AI-accelerated uplink proof-of-concept (AI Transmit Diversity) — was an official AI-RAN Alliance demonstration at MWC 2026 (validated with Keysight) that delivered up to 30–40% cell-edge uplink gains, framed within a Model Lifecycle Management (LCM) architecture for AI in the air interface. Six years across 3GPP RAN1 / RAN2 standardization (PHY-to-RRC), PHY-to-MAC system-level simulation, and on-device machine learning deployed to modem MCU / DSP — from quantization-aware training and model compression to live silicon. IEEE-published in 5G / NB-IoT system prototyping (53 citations, h-index 4); co-inventor on an allowed US patent (US/EP/CN family). Currently building agentic AI for system-level simulation and ray-tracing channel modeling at the AI-RAN frontier.
02
Highlights
- AI-RAN Alliance @ MWC 2026 — sole owner of the UE-side AI proof-of-concept end-to-end: model training → MCU/DSP embedded implementation → live smartphone; exhibited at both the MediaTek and AI-RAN Alliance booths, validated with Keysight.
- Patents — two patent families as co-inventor (assignee MediaTek): a US relay/multipath patent (US 2024/0064605 A1) at Notice of Allowance with EP + CN counterparts, plus a 2026 AI-uplink filing.
- Research record — 4 IEEE publications (DSP, GLOBECOM, INFOCOM), 53 citations, h-index 4; upstream open-source contributor to OpenAirInterface (OAI).
- Full-stack span — from 3GPP PHY-to-RRC and system-level simulation, to independently implementing an IAB/Relay framework and prototyping eNB/gNB across 4G LTE, NB-IoT, and 5G, to deploying learned models on modem MCU/DSP.
03
Professional Experience
Senior Engineer, MediaTek
Communication System Design — Advanced Communication Technology (R&D)
- AI TxD — AI-RAN @ MWC 2026: Designed and delivered the UE-side proof-of-concept for “AI Transmit Diversity and Model Lifecycle Management,” exhibited as “AI for 6G Uplink.” An “AI-enabling-AI” approach combining RAN-assisted AI decision-making, cloud-based site-specific model re-training, and over-the-air (OTA) model updates — yielding measurable gains in uplink throughput, spectral efficiency, and reliability. Owned the UE-side pipeline end-to-end: training → embedded MCU/DSP implementation → live smartphone.
- Model Lifecycle Management (LCM): Defined the LCM architecture for AI in the air interface — site-specific re-training + OTA update — so deployed models sustain performance across changing network environments, extending the work from a single model to a deployable framework.
- Proprietary device AI: Built AI solutions for UE across mobility, radio resource management (RRM), and UE transmission (TX), pairing signal-processing domain knowledge with on-device ML under tight latency and power budgets.
- AI-RAN frontier (current): Developing agentic AI to advance a system-level simulator, and working with ray-tracing channel modeling to generate site-realistic training data for air-interface AI.
Engineer, MediaTek
Communication System Design — 5G NR Research (R&D)
- System-level simulation: Built and extended a 5G NR system-level simulator modeling gNB–UE interaction, with full PHY-to-MAC modeling (channel, PHY, MAC) derived directly from 3GPP technical reports and contributions (TR/Tdoc).
- 3GPP standardization & patents: Worked across RAN1 (physical layer) and RAN2 (L2/RRC); co-invented a PDCP-layer multipath relay patent (US 2024/0064605 A1 — allowed on first office action; family US/EP/CN).
- Relay / IAB & prototyping: Independently implemented an IAB/Relay framework and prototyped eNB/gNB base stations across three RATs — 4G LTE, NB-IoT, and 5G — with command of the full 3GPP stack from PHY to RRC.
Research Intern, EURECOM
OpenAirInterface (OAI) — NB-IoT development team
- Built the nFAPI-standard IF module enabling the MAC–PHY functional split; shipped into OAI’s
developanddevelop-nb-iotbranches (LTE / NB-IoT / 5G). - Part of an open-source 3GPP Release-13 NB-IoT eNB collaboration with EURECOM, b<>com, and Nokia (France) — foundation of a sustained NB-IoT line of work later published across IEEE GLOBECOM, INFOCOM, and DSP.
Graduate Research, NTUST — Wireless Systems Lab
Advisor: Prof. Ray-Guang Cheng
- M-CORD — Project Leader: Led planning and design to virtualize an OAI base station into an XOS service on M-CORD using SDN/NFV (OpenStack, ONOS), with NCTU and US ON.Lab.
04
Patents
US 2024/0064605 A1 — Notice of Allowance, 2026
Method of Path Selection in PDCP Layer to Support Multipath Configuration. Path selection for UE-to-network relay using indirect-path side information to satisfy uplink QoS, at per-radio-bearer / per-QoS-flow / per-logical-channel granularity. Co-inventor; assignee MediaTek Inc.
US Patent Application — Serial 19/667,845 — filed 2026
Methods and User Equipment (UE) of Single-Layer Uplink Transmission. Co-inventor; assignee MediaTek Inc. Related to the AI-accelerated uplink work demonstrated at MWC 2026.
05
Selected Publications
Design and Prototype of a Virtualized 5G Infrastructure Supporting Network Slicing
Prototyping of Open-Source NB-IoT Network
Design and Implementation of an Open-Source NB-IoT eNB (Demo Abstract)
Open NB-IoT Network in a PC
Full record & citations on Google Scholar — 53 citations · h-index 4.
06
Technical Skills
AI / ML
On-device inference (MCU/DSP)Model training & deploymentQuantization-aware trainingModel compressionEdge optimizationModel Lifecycle Management (LCM)Agentic AITime-series prediction
Wireless / 3GPP
5G NR5G-Advanced6GNB-IoT4G LTEPHY-to-RRC stackRAN1 / RAN2TR / TdocIAB / RelaySystem-level simulation
Tools / Platforms
Ray-tracing channel simulationOpenAirInterface (OAI)SDR eNB/gNB prototypingSDN / NFV (OpenStack, ONOS)
Programming
CC++PythonDSP / embeddedLinuxGit