T
Tensordyne Tech Stack
Custom silicon and system software for low-power AI inference at scale
Tensordyne designs and manufactures custom AI inference accelerators using proprietary logarithmic math to replace multiply operations with additions, reducing power consumption at the algorithmic level. The tech stack is deeply hardware-focused (Verilog, SystemVerilog, Cadence, RISC-V, ASIC workflows) with an engineering-heavy org (11 of 14 active hires) split between chip design and systems integration, signaling active bringup of silicon and deployment platforms.
Tech Stack 28 technologies
What Tensordyne Is Building
◆Challenges
- Accelerating multimodal generative ai at scale
- Low-power generative ai inference
- High-performance generative ai inference
- Inefficient eda workflows
- Scalable devops systems
- Selling into hyperscalers
- Complex multi-stakeholder sales cycles
- Developing world-class ai inference compute systems
- High-volume manufacturing
- Complex hardware system
▲Active Projects
- Design and bringup of development systems
- Next-generation ai processors for generative ai inference acceleration
- Reference platforms
- Volume products
- Next generation ai inference compute products
- Deployment of inference systems
- Ai inference hardware products
- Evaluations and pocs
- Development of devops systems for asic workflows
- Optimization of eda tools for asic chip design
Hiring Activity
15 roles · 10 in 30d
Department
Seniority
Company intelligence
