Deep Learning Compiler Engineer
Quadric, Incover 2 years ago
Pune, IndiaSenior
Responsibilities
- Drive the lowering and optimization of deep neural networks using Quadric technology.
- Develop mathematical and algorithmic optimizations for NP-hard problems.
- Collaborate with the software team on graph-based execution algorithms for the Quadric architecture.
- Develop optimization passes for efficient lowering of deep learning and high-performance computing algorithms on the Quadric EPU.
- Build a deep understanding of the hardware platform and low-level software to optimize application performance.
Requirements
- MS or Ph.D. in Computer Science or a related field with a minimum of eight years of industry experience.
- Strong background in numerical and/or algorithmic optimization.
- Understanding of application-appropriate heuristics for NP-hard problems.
- Knowledge of classical and machine learning algorithms, including computer vision, DSP, and deep neural networks.
- Strong background in graphs and related algorithms.
- Proficiency in C++ version 11 or later is preferred.
- Experience using or developing in TVM or MLIR is preferred.
- Knowledge of compiler front-end and back-end techniques is preferred.
Benefits
- Competitive salary and meaningful equity.
- Opportunity to join a small, rapidly growing team and make an immediate impact.
- Opportunity to build long-term career relationships.
- Politics-free community and an environment supporting personal and professional relationships.
- On-site position in Pune, India.
Tech Stack
Categories
About Quadric, Inc
Quadric licenses an AI processor architecture—the Chimera GPNPU—that runs neural-network inference and conventional C++ code on a single programmable core for edge devices. Its licensable IP targets automotive, industrial, robotics, and embedded systems, enabling customers to integrate a unified accelerator/processor into their SoCs. Founded in 2017 and headquartered in Burlingame, California, the privately held company was started by technologists from MIT and Carnegie Mellon.