Principal Software Engineer

Overview

Pune, Maharashtra, INঅফিসে

Responsibilities

Own, maintain, and enhance the end-to-end automation flow used to:
Compile neural networks using the XNNC compiler.
Run generated RTL binaries on Palladium Z3 emulation.
Drive Joules gate-level power analysis.
Develop and maintain the Python, CSH, and Bash scripts that integrate and orchestrate the flow.
Add support for new AI network configurations, hardware accelerator variants, reference builds, and EDA tool revisions.
Maintain current documentation, patches, reference configurations, and reproducible build instructions.
Identify, debug, and resolve automation, infrastructure, and tool-integration issues across the workflow.

10+ years of experience in various embedded software or general software development roles.
3+ years of experience with AI frameworks, AI network compilation, or related tooling.
Experience with one or more of: TensorFlow, PyTorch, ONNX, Apache TVM, or AI network compilers.

Bachelor’s or Master’s degree in Software Engineering, Electronics Engineering, Electrical Engineering, Computer Engineering, or a related discipline.

Experience using AI-assisted development tools for rapid prototyping and effective technical prompting.
Experience with Cadence tools, including Palladium and Joules, or equivalent EDA/emulation/power-analysis workflows.
Familiarity with IP design platforms, RTL simulation or emulation, gate-level power analysis, and AI accelerator validation.
Strong written communication skills for producing technical documentation, runbooks, and performance/power reports.

Generate complete performance and power collateral for each neural network processed through XNNC.
Collect and summarize:
Cycle count, frames per second (FPS), and memory-bandwidth metrics from XTSC simulation.
End-to-end performance metrics from Palladium emulation.
Dynamic power estimates, including total and memory power, from Joules gate-level analysis.
Deliver results consistently across reference builds in standard formats, including cycle-count CSVs, power CSVs, and consolidated summary reports.
Validate result quality, consistency, and reproducibility before collateral is shared with stakeholders.

Serve as the primary internal point of contact for users of the COLLATERAL-AUTO project.
Onboard new users and provide guidance on running the XNNC, Palladium, and Joules workflow.
Debug user runs and triage issues related to environments, scripts, CAD/EDA tools, and flow configuration.
Review performance and power summaries for correctness, completeness, and consistency.
Improve team-wide productivity by creating reusable guidance, troubleshooting documentation, and reliable reference workflows.

Strong Python scripting skills.
Proficiency with CSH and Bash shell scripting.
Working knowledge of JSON and automation-oriented configuration formats.
Understanding of hardware edge-AI accelerators, RTL execution flows, and performance analysis.
Familiarity with debugging complex software, toolchain, and multi-stage automation issues.