Computing, Electronics and Digital Systems
Work description
? Review the current state of the onnx-flow tool, the ONNX format, and mapping tools for CGRA; ? Implement new transformations that expose explicit scalar computation and reduce the complexity of composite nodes (e.g., MatMul, Conv, ReduceSum), with a focus on merging, removing redundant loops and coalescing dimensions; ? Create realistic examples of transformed ONNX models, and validate the mappability of the generated graphs with existing mapping tools; ? Collaboration in the writing of a scientific article to disseminate the results.
Academic Qualifications
? Degree or enrollment in a master's degree in electrical engineering, computer science, or a related field;
Minimum profile required
? Programming experience (ideally JS/TS)? Fluency in English (written and spoken)
Preference factors
? Experience with structuring or manipulating graphs (e.g., ASTs, DFGs) ? Interest in compilers and/or familiarity with the ONNX format ? Fluent in Portuguese and English (written and spoken)
Application Period
Since 24 Jul 2025 to 06 Aug 2025
Centre
Human-Centered Computing and Information Science