Yuan-Pin Huang
Executive Summary
MediaTek E9 / Staff Engineer with 10+ years of experience spanning ASIC RTL design, SoC/IP integration, FPGA/CPLD development, full-SoC emulation, physical-aware implementation, and design-driven silicon productization. Defines subsystem architecture boundaries, integration contracts, closure criteria, and pre-silicon de-risk plans across RTL, DV, DFT, APR, emulation, and post-silicon teams. Owned AXI-based SRAM controller subsystems across advanced-node modem SoCs, integrated 112G retimer/MACsec IP and PCIe Gen5 test-chip IP, and drove HAPS/Palladium-class validation milestones to reduce bring-up risk. Strong fit for Staff roles requiring architecture definition, technical risk review, cross-team sign-off readiness, and execution from pre-silicon definition through production ramp.
Core Competencies
Professional Experience
- Architecture definition & review: Defined SRAM controller integration boundaries, AXI3 transaction behavior, 8-way interleave assumptions, low-power
SLEEPBpolicy, CDC/DFT/MBIST expectations, and timing-constraint handoff criteria with SoC, DV, DFT, and APR stakeholders. - Cross-team closure framework: Established practical closure checkpoints for RTL quality, CDC cleanliness, MBIST/DFT insertion readiness, synthesis constraints, APR handoff, emulation milestones, and post-silicon debug coverage to keep subsystem risks visible before tape-out and bring-up.
- Memory subsystem ownership: Owned AXI3-compliant SRAM controller subsystems across MT6980 / MT6990 / MT2737 / MT6986 / MT6988 / MT2739 modem programs, covering RTL design, controller implementation, verification, Lint/CDC, MBIST/DFT insertion, synthesis, clock-tree constraints, and APR timing-closure collaboration.
- 1408KB SRAM controller: Designed and delivered a single-port TSMC SRAM controller with 8-way interleave and automatic low-power
SLEEPBmanagement; closed timing at 728 MHz for a 200K+ cell, 350K μm² subsystem on N4 modem products. - 512KB SRAM controller: Previously delivered AXI3-compliant single-port SRAM controller at 533 MHz, covering 50K+ cells and 175K μm² area, with full RTL-to-implementation ownership.
- 112G retimer / MACsec sign-off readiness: Integrated Rambus MACsec subsystem and defined front-to-backend closure checkpoints for DFT insertion, synthesis, clock constraints, congestion/timing risk, and APR handoff; coordinated with APR stakeholders to achieve timing closure at 1.308 GHz on a 5M+ cell, 3.5M μm² block.
- Emulation de-risk plan: Translated architecture and bring-up risks into HAPS/Palladium validation milestones for modem and retimer projects; connected four HAPS-80 systems through HT3 cables for whole-chip verification and system-level debug readiness.
- Bring-up risk reduction: Early-ported low-power behavior into FPGA emulation to validate power-state assumptions before silicon; integrated thermal controller logic and coordinated sensor-position planning with SoC and physical-design stakeholders.
- Design-aware productization readiness: Converted design intent into production test readiness by driving Day-0 coverage/pass-rate tracking, JTAG/UART/GPIO testability verification, probe card/load board capability definition, test-flow/program definition, yield/DPPM tracking, and MP flow/program release for MT6986 / MT2739 and TPU v7 programs.
- PCIe Gen5 integration contract: Integrated Synopsys PCIe Gen5 MAC into ASIC test-chip environment and aligned MAC/PHY/DV/APR interfaces, Lint/CDC expectations, MBIST insertion scope, local VIP setup, and PHY/MAC co-verification criteria.
- Cross-functional closure: Partnered with DV, PHY, BE, and APR teams to close verification and implementation risks, pass PHY+MAC test plans, and reach timing closure at 1 GHz.
- Emulation enablement: Supported ASIC project emulation tasks and helped build Synopsys FPGA environment components for faster system-level validation and debug.
- Server control logic: Developed CPLD/FPGA logic for server power control, LED control, debug interfaces, I2C/SPI connectivity, and BMC communication on server boards.
- Debug architecture: Architected and implemented an OCP-compliant debug card using AVR microcontrollers, improving engineering diagnostics and platform bring-up efficiency.
- FPGA AI demonstration: Built Intel Altera FPGA-based server platform capabilities for AI training demos using OpenVINO / TensorFlow workloads, including CNN, GNN, and reinforcement-learning demonstrations for Computex.