Sr. Engineer, AI Compilers
Tenstorrent AI ULC (2025-present)
Contributed to the design and implementation of MLIR-based AI compilers for Tenstorrent hardware, such as blackhole and wormhole using C++20 and Python:
- Pushed the generality of the compiler and added support in the optimizer for more operators and graphs. This automation resulted in significant reduction in developers' time.
- Worked on ttnn-jit frontend and made it pip install-able for easier usage.
- Developed graph-based approaches for generating ttnn IR from a decorated python function to be used in ttnn-jit. This added support for TMs and reduction ops, as well as dynamically decomposing composite ops.