Lead SoC Modeling Engineer

Lead SoC Modeling Engineer
Location: Bengaluru
Experience: 8+ Years
Role Overview:
We are seeking a highly skilled Lead SoC Modeling Engineer to develop and validate bit-accurate reference models for next-generation AI/ML and high-performance compute SoCs. The role requires deep expertise in C++ modeling, floating-point arithmetic, quantization techniques, and Model-RTL correlation to ensure functional correctness and first-pass silicon success.
Key Responsibilities:
Develop and maintain bit-accurate and cycle-accurate C++ reference models for complex SoC and AI accelerator subsystems.
Drive Model-RTL correlation, debug mismatches, and ensure architectural compliance.
Implement and validate numerical formats including FP32, FP16, FP8, BF16, INT8, and INT4 quantization schemes.
Collaborate with Architecture, RTL Design, and Verification teams to define modeling requirements and validation strategies.
Create validation frameworks, test vectors, and automated regression environments.
Analyze numerical accuracy, corner cases, and performance characteristics across compute pipelines.
Enhance modeling methodologies and automate flows using Python and scripting languages.
Provide technical leadership and mentor junior engineers.
Technical Skills:
Strong expertise in modern C++ and object-oriented programming.
Hands-on experience with bit-accurate/cycle-accurate modeling and Model-RTL correlation.
Deep understanding of floating-point arithmetic and quantization formats (FP8, BF16, INT8, INT4).
Strong knowledge of AI/ML accelerators, matrix operations, vector processing, and computer architecture.
Experience with SystemVerilog/UVM-based verification environments is desirable.
Proficiency in Python and Linux development environments.
Familiarity with SIMD architectures, performance modeling, and debug methodologies is a plus.
Qualifications
B.E./B.Tech/M.E./M.Tech in Electronics, Computer Science, or related disciplines.
Proven experience delivering high-quality reference models for complex SoCs, CPUs, GPUs, or AI accelerators.
Show more Show less