About the role:As a member of Tensordyne's ASIC team, you will be responsible for the microarchitecture and design implementation of a high-performance and low-power convolutional neural network accelerator ASIC that forms the core of the company's flagship perception module product for autonomous driving applications.This ASIC's design closely couples novel computational accelerator units with 3rd-party SoC IP blocks to form an end-to-end vision perception module that achieves record-breaking computational performance at low power.
Your responsibilities will be wide-ranging and run the gamut of microarchitecture design and documentation, Verilog RTL implementation, selecting and integration 3rd-party IP blocks, working closely with design verification engineers, running frontend synthesis and working with a backend team on place-and-route guidance and timing closure.
Responsibilities:- Author/Understand micro-architecture specifications and participate in specification and test plan reviews.
- Implement complex RTL designs.
- Integrate CPU and other relevant IPs into the CPU sub-system.
- Collaborate with the physical design team to resolve implementation and timing issues and to optimize power.
- Analyze code coverage and provide feedback to the verification team to achieve coverage closure.
- Perform diagnostic and post-silicon validation tests, as well as assist with software bring-up in the lab.
Required Qualifications:- 3+ years of ASIC design experience with a demonstrable track record of RTL logic design in multi-million gate ASICs with Verilog or System Verilog.
- Experience in IP integration, specifically CPU IP into SoC.
- Knowledge of ARM/RISC-V/MIPS Architectures, Memory hierarchy, Cache coherency, Virtual memory, Multicore CPU operation
- Familiarity with AMBA/APB/AXI Protocol
- Familiarity with processor peripheral interfaces like SPI, eMMC, *MII, GPIO, I2C ....
- Hands-on experience in all aspects of the ASIC development process with proficiency in front-end tools and methodologies.
- Familiarity with low power design. UPF flow for defining power intent of chips with multiple power domains
- Previous experience with timing closure at high frequencies is a plus.
- Interest to explore AI architectures for convolution, transformer and other kinds of workloads
- Self-starter and highly-motivated to work in a dynamic start-up environment.
- B.S. (M.S. preferred) degree in Electrical or Computer engineering.
