Skip to main content
Nithish Reddy

Nithish Reddy

Project Engineer

Talent
BengaluruMember since August 17, 2026
C++PythonPyTorchMachine learningDeep learningONNXMLIRLLVM
·

Hire this person

Send a job offer directly to this candidate

C++PythonPyTorchMachine learningDeep learning

About

Developed MLIR compiler passes for custom AI Accelerator, GEMM tiling based on the systolic array sizing, VPU tiling, layerize pass to combine the operations and create a region, custom backend dialect (vendor dialect), passes and instruction generation.

Developed

Automatic mixed precision Quantization passes at compiler level (specifically at the linalg level, based on the precision policy) and its instruction generation. Developed production level compiler frontend for the pytorch to generate the common IR, which include maintaining state across frameworks, frontend contract to be followed across frameworks specific to the backend.

Experience

Project Engineer at CDAC Bangalore (2025-07 – Present)

  • Developed MLIR compiler passes for custom AI Accelerator, GEMM tiling based on the systolic array sizing, VPU tiling, layerize pass to combine the operations and create a region, custom backend dialect (vendor dialect), passes and instruction generation.
  • Developed Automatic mixed precision Quantization passes at compiler level (specifically at the linalg level, based on the precision policy) and its instruction generation.
  • Developed production level compiler frontend for the pytorch to generate the common IR, which include maintaining state across frameworks, frontend contract to be followed across frameworks specific to the backend.
  • Developing the training and inference workflows for an AI Accelerator, working deeply with model internals including implementation by frameworks, computation graphs, operators, and data flow.
  • Enabling inference support for popular deep learning frameworks including PyTorch, TensorFlow, JAX and other frameworks Using the ONNX exporter.

Machine Learning Intern at KOSKILL (2023-01 – 2023-03)

  • Developed SkimLit system for automated medical abstract sentence classification using multiple ML architectures (Bi-LSTM, transformers)
  • Achieved high classification accuracy on PubMed 200k dataset by implementing hybrid embedding approaches and model optimization techniques

Education

BTech in ECE – National institute of technology goa

Intermediate – Narayana junior college

– vijetha high school

Skills

Project Engineer

Reviews

Related