Architect Labs · San Francisco
Hardware & ElectricalHardwareFullTime
Architect is a frontier AI lab for custom silicon. We partner with frontier labs, clouds / neoclouds, physical AI companies, and advanced fabs to tape-out custom chips co-designed for next-generation AI workloads. Our goal is to compress end-to-end software to silicon timelines, and maximize intelligence per watt and per dollar for the world. We are a small exceptional team across silicon, systems, software and frontier AI. Our team have led research teams at nearly every frontier AI lab, and at some of the most complex SoCs in the world.
As a Founding Member of the Technical Staff on the RTL Design team at Architect, you’ll own the AI-driven microarchitecture and RTL design of the on-chip interconnect fabric and high-speed I/O data movement subsystems going into production silicon. You will define, drive, and revise the block-level micro-architecture specification for NoC routers, crossbar switches, high-speed fabric bridges, and peer-to-peer data transfer engines — ensuring low-latency, high-bandwidth, and deadlock-free communication across all SoC agents.
Own the on-chip fabric RTL end-to-end: from NoC topology and router microarchitecture through code generation, lint, CDC, synthesis, and timing closure using our AI-driven design flow.
Design and implement AMBA-based interconnect fabrics: including AXI/ACE/CHI-compliant crossbar switches, network interfaces (NIs), protocol converters (AXI-to-CHI bridges, AXI-to-AHB/APB downconverters), and multi-layer interconnect configurations optimized for ML accelerator traffic patterns.
Architect NoC routers and topologies: including virtual-channel routers, wormhole/flit-based switching, adaptive routing algorithms, QoS-aware arbitration (bandwidth regulation, latency-critical path prioritization), and deadlock-free network design for mesh/ring/tree topologies.
Design high-speed I/O fabric bridges and peer-to-peer engines: including PCIe/CXL-to-fabric bridges, chip-to-chip interconnect logic (UCIe, custom die-to-die links), peer-to-peer DMA controllers for direct device-to-device transfers bypassing host memory, and coherent/non-coherent multi-chip fabric extensions.
Work directly with the principal architect to refine microarchitectural specs, resolve implementation trade-offs (latency vs. bandwidth vs. area, coherence overhead vs. performance), and feed area/timing/power realities back into the architecture and internal AI systems.
Define and maintain interface specifications: AMBA AXI4/AXI5, ACE/ACE-Lite, CHI (with snoop filter interfaces), AXI-Stream for streaming datapaths, custom sideband channels for QoS/ordering, and high-speed SerDes-facing interfaces for off-chip links.
Build and maintain RTL infrastructure for our in-house AI-driven flow: design automation scripts, NoC configuration generators, regression flows, lint/CDC waivers, and integration collateral for the interconnect subsystem.
Close collaboration with DV: Support verification bring-up with interconnect reference models, protocol compliance checkers (AXI/CHI protocol monitors), SVA assertions for ordering rules and deadlock freedom, coverage plans targeting corner-case traffic scenarios (multi-master contention, QoS starvation), and architectural documentation for verification closure.
Close collaboration with SW and ML: Support and guide our SW and ML experts to revise and improve our in-house AI flow based on your interconnect domain expertise — particularly around traffic modeling and fabric configuration optimization.
Support FPGA prototyping on Xilinx for early functional validation of the fabric, including multi-master traffic generation and performance characterization on FPGA platforms.
Degree: Bachelor’s, Master’s, or PhD in Electrical Engineering, Computer Engineering, or a closely related field.
Experience: 5+ years (10+ preferred) in RTL design with at least one advanced-node tapeout experience involving on-chip interconnects, NoC fabrics, or high-speed I/O subsystems.
AMBA Protocol Expertise: Deep familiarity with ARM AMBA protocol suite — AXI4/AXI5 (channel mechanics, burst types, ordering, exclusive access), ACE/ACE-Lite (coherence transactions, snoop channels), CHI (request/response/data/snoop flits, home nodes, snoop filters), and legacy AHB/APB for peripheral integration.
NoC/Fabric Design: Hands-on experience designing or owning crossbar switches, NoC routers, or multi-layer interconnects including arbitration schemes (round-robin, priority, age-based, bandwidth-regulated), virtual channel management, flow control (credit-based, ready/valid), and QoS mechanisms.
High-Speed I/O Integration: Experience with HSIO bridge logic — PCIe root complex/endpoint bridge design, CXL.io/CXL.mem/CXL.cache protocol translation, or custom chip-to-chip links (UCIe, proprietary die-to-die) including link-layer protocols, credit management, and replay/retry logic.
Peer-to-Peer Data Movement: Experience with peer-to-peer DMA architectures, zero-copy data transfer engines, scatter-gather descriptors, and direct device-to-device communication paths that bypass host memory bottlenecks.
SystemVerilog: Clear, synthesizable, lint-clean RTL with strong design habits — parameterization for configurable port counts and data widths, modularity for hierarchical fabric composition, and configurability for different topology and QoS instantiations.
SoC Methodology: Solid grasp of synthesis, timing constraints (especially for wide crossbar paths and high-radix switches), clock domain crossings (fabric-to-IP clock boundaries, async bridge design), reset strategies, and power management for interconnect logic.
Python: Strong skills for design automation, traffic generation/analysis, NoC configuration scripting, regression infrastructure, and tooling.
PPA Ownership: Experience taking an interconnect or fabric block from RTL through synthesis and working with PD teams on timing/area/power closure — particularly for wide-datapath crossbars and high-frequency router pipelines.
Experience with coherent multi-chip/multi-die interconnect architectures (chiplet-based designs, UCIe, BoW).
Familiarity with hardware coherence protocols: MOESI/MESIF state machines, snoop filter design, directory-based coherence.
Experience with network-on-chip research: adaptive routing, congestion management, topology optimization, or formal deadlock analysis.
Low-power design techniques for interconnect: clock gating idle ports, power gating unused links, link-level power states (L0s/L1), dynamic frequency/width scaling.
FPGA prototyping experience (Xilinx Vivado/Vitis), especially with AXI interconnect IPs, NoC IPs (Versal), or custom fabric implementations.
SVA assertions for protocol compliance: AXI ordering rules, CHI transaction flows, deadlock detection, and livelock/starvation monitors.
Prior IP building and delivery experience for NoC IPs, AXI interconnect IPs, PCIe controllers, or CXL endpoint/switch IPs.
Performance modeling: experience building or using NoC simulators (e.g., BookSim, Garnet) or system-level traffic models to validate fabric microarchitecture.
Domain-specific research contributions: publications or patents in on-chip networks, interconnect architectures, or high-performance data movement for ML/HPC workloads.
You’ll join a founding team building the future of chip design at the intersection of AI and silicon. Your interconnect and fabric expertise will directly shape the communication backbone of production ASICs — enabling the data movement performance that ML workloads demand — and influence how AI transforms hardware development from spec to tapeout.