D

d-Matrix

Delivers memory-integrated AI compute platforms

Staff Analog Design Engineer - Serdes Clocking

Silicon & ASIC DesignConfirmed live in the last 24 hours
Full-timeSanta Clara, CA, USA$155k - $250k /yrHybrid

About the Job

Requirements

  • A Bachelor of Science, Master of Science, or PhD in Electrical Engineering and at least 5 years of hands-on analog and mixed-signal circuit design experience.
  • Demonstrated ownership of a low-jitter phase-locked loop, phase interpolator, or clock and data recovery loop through silicon in a high-speed SerDes link.
  • Deep knowledge of phase-locked loop architecture, loop dynamics, voltage-controlled oscillator design, and phase-noise theory, and working knowledge of clock and data recovery architectures and the loop analysis behind jitter tolerance and jitter transfer.
  • Practical experience with jitter decomposition and budgeting methodology, including quantifying contributors in simulation.
  • Circuit design experience in advanced FinFET or gate-all-around process nodes, using industry-standard analog design and simulation tools and methodologies such as Cadence Virtuoso, Spectre, Monte Carlo/corner analysis, and post-layout simulation.
  • Silicon bring-up and bench-debug experience with laboratory equipment such as phase-noise analyzers, high-speed oscilloscopes, and bit error-rate testers.

Responsibilities

  • Own transistor-level design of low-jitter frequency synthesis using LC and/or ring phase-locked loops, including loop filters, charge pumps or digital loops, voltage-controlled oscillators, and divider chains.
  • Own the phase interpolator and phase-rotation scheme, including integral and differential nonlinearity, glitch behavior, and interpolation resolution against clock and data recovery requirements.
  • Design high-frequency clock distribution across the SerDes macro, including buffer chains, duty-cycle correction, quadrature generation and correction, skew control, and supply isolation.
  • Define and verify clock and data recovery architecture and loop dynamics, including bandwidth, jitter tolerance, jitter transfer, lock acquisition, and frequency-offset tracking; model loop behavior in MATLAB, Python, or equivalent; correlate models against transistor-level and silicon results in coordination with datapath and digital signal processing owners.
  • Build phase-noise and jitter budgets for the link and own analysis of jitter sources.
  • Build verification environments for the blocks, including periodic steady-state/harmonic-balance and transient-noise analysis, process-voltage-temperature corner coverage, Monte Carlo analysis, and post-layout back-annotated simulation; provide behavioral and Verilog-A models for full-chip verification.
  • Drive layout of the blocks in deep sub-micron nodes, with particular attention to inductor and voltage-controlled oscillator layout, shielding, symmetry, and supply/substrate isolation.
  • Take the blocks through silicon bring-up and bench characterization, measuring phase noise, jitter tolerance, and lock range, and correlate measured results against simulation.

Desired Qualifications

  • Die-to-die interconnect experience with UCIe, BoW, or a proprietary die-to-die physical layer, including clocking architecture choices for short-reach, low-energy-per-bit links.
  • Experience with forwarded-clock and matched-source-synchronous architectures alongside embedded-clock clock and data recovery.
  • Experience with digital and hybrid phase-locked loop/clock and data recovery implementations, including their digital calibration.
  • Familiarity with relevant specifications, including UCIe, OIF CEI-112G/224G, IEEE 802.3, and PCIe Gen6/Gen7.
  • Experience porting a clocking block across process nodes, with a realistic understanding of the effort involved.
  • Participation in or contribution to standards bodies.

About the Company

d-Matrix

Delivers memory-integrated AI compute platforms

d-Matrix provides scalable AI compute platforms for large datacenters with a focus on energy efficiency. Its core offering, the digital in-memory compute (DIMC) engine, places compute directly in programmable memory and uses a fabric of low-power, modular chiplets plus software to optimize AI inference. It differentiates itself by combining memory-integrated computation with a modular, chiplet-based design that reduces power use and data movement for high-volume workloads. The goal is to help large datacenters run AI workloads more efficiently at scale through customizable, energy-efficient compute solutions.

Company size
201-500
Company stage
Series C
Total funding
$429M
Headquarters
Santa Clara, CA
Founded
2019

Recently Posted Jobs

Explore more opportunities at d-Matrix

See more jobs