Principal Microarchitect

Acceler8 Talent is seeking a Principal Micro-Architect (Compute Engine) to join an early stage startup backed by a Tier-1 VC that is rethinking AI infrastructure from first principles.

Founded by highly respected industry veterans, they are innovating at the chip and system level to deliver an order of magnitude better performance-per-watt for inference, which would mean a huge economic shift for anyone running large scale models, unlocking larger context windows and longer generations, making new workloads economically viable.

Responsibilities:

 ● Design execution pipelines, scheduling, and control logic

 ● Define datapaths, register files, scratchpads, and local memory hierarchies

 ● Implement support for modern numeric formats including FP16, BF16, FP8, & FP4

 ● Optimize for performance, utilization, power, and area

 ● Balance compute throughput with memory bandwidth and SoC constraints

 ● Work closely with architecture, compiler, kernel, and performance teams

 ● Analyze and optimize select AI workloads

 ● Evaluate ISA / programming model trade-offs

 ● Use performance modeling and profiling to guide architecture decisions

 ● Lead initial RTL implementation

 ● Support ongoing RTL implementation, verification, and timing closure

 ● Participate in emulation, debug, bring-up, and post-silicon performance tuning

Qualifications:

 ● Strong background in AI compute or accelerator design as evidenced by design of complex units with multiple blocks over multiple generations of process technology

 ● Deep understanding of:

• Matrix and vector pipelines
•  Floating point numerics
• Quantization
• Performance per watt optimization

 ● Experience delivering real silicon

 ● Experience in startup or early-stage silicon programs is a plus

 ● MS or PhD and 3-5 years of experience, or equivalent

If you're looking for massive ownership, huge impact, and the opportunity to build from the ground up, please apply here or reach out to me at ltomaszko@acceler8talent.com to hear more.