Title: Chief Compiler Architect
Where: Remote work is fine
Job Description:
Vathys is building a fundamentally novel, programmable, parallel processors.
At core, the architecture can be thought of as a static dataflow processor.
As such, compilation consists of extracting a dataflow graph from the program and
mapping it to the dataflow engine in the Vathys processor.
We are taking a refreshingly simple yet effective approach to compiler design.
The goal is to create a basic compiler which is the pretty much the simplest possible
compiler, and then, to use machine learning techniques as the optimizer. This way, there is little
effort needed on the optimizer components of the compiler. This means in part that considerable effort
will be needed on the frontend.
This also helps hide the considerable complexity of traditional place and route optimizers which are often
necessary for other dataflow processor compiler toolchains.
Skills and Qualifications:
1. Experience in developing compilers (must have).
2. Experience in machine learning, especially deep learning and reinforcement learning (good to have).
3. Experience in working with computational graph intermediate representations, such as Glow, ONNX (very good to have).
4. Experience with auto-tuners, especially domain specific auto-tuners like TVM is good to have
Compensation: Generous compensation will be given to the right talent.
If you're interested, please email tapabrata_ghosh (a t) vathys.ai
