Publications

Neuromorphic Intermediate Representation for T1C Hardware

Feb 2026 ยท pypi.org/project/t1c-ir

Authors: Murat Isik, Ayesh Ahmad, Subham Das, Anurag Daram

  • Context: Developed an intermediate representation (IR) for compiling and deploying neural networks onto Type 1 Compute (T1C) neuromorphic hardware, with first-class support for hybrid SNN/ANN graphs.
  • Methods: Specified IR semantics and a primitive set relative to the Neuromorphic IR (NIR) baseline; implemented a reference stack in Python and C++ with PyTorch-based authoring/export and HDF5-backed artifact exchange.
  • Contributions: Led research and engineering from IR design through implementation; introduced additional primitives beyond NIR for hybrid and hardware-specific workflows; shipped the open-source t1c-ir package on PyPI.