Senior Computer Vision Library Developer

GSI Technology is a leading international company publicly traded on NASDAQ, specializing in the development of the Gemini® processor — a cutting-edge Associative Processing Unit (APU) designed for computer-in-memory acceleration.

We are seeking a highly skilled and self-driven Senior Software Engineer to lead the development and optimization of large language model (LLM) implementations on a proprietary Associative Processing Unit (APU). This role combines high-level machine learning understanding with low-level system and performance engineering, primarily using C and C++. The ideal candidate will possess deep knowledge of transformer architecture, be capable of writing reference implementations, debugging Python pipelines, and optimizing memory and performance across the full stack.

Key Responsibilities:

• AI Library Development:
• Build and optimize software libraries for Large Language Models (LLMs) and Large Vision Models (LVMs).
• System Architecture:
• Design end-to-end system flows integrating AI models, particularly focusing on computer vision applications.
• Hardware Optimization:
• Improve performance under hardware constraints (e.g., latency, memory, compute limitations).
• Cross-Stack Integration:
• Ensure seamless interaction between AI models, software infrastructure, and underlying hardware.

Required Skills:

At least 5 years of experience in:

✔️ LLM/LVM Development: Practical experience with transformer architectures and vision-language models

✔️ Computer Vision: Strong background in CV pipelines and multimodal systems

✔️ System Design: Proven ability to architect complex software systems from concept to deployment.

Technical Specializations

✔️ Hardware-aware optimization techniques (quantization, pruning, kernel fusion)

✔️ Performance profiling tools (e.g., PyTorch Profiler, NVIDIA Nsight)

✔️ Low-level optimization (CUDA, OpenCL, or hardware-specific SDKs)

Preferred Qualifications

• Experience with AI accelerator architectures (GPUs, TPUs, NPUs)
• Familiarity with model compression techniques and distributed training
• Background in ML compiler frameworks (TVM, MLIR, or XLA)

Our Privacy Policy: Your resume and information will be kept confidential.