AI Inference Engineer - Model Optimization & Deployment

📋 Description

• Optimize large-scale models (LLMs/VLMs) with quantization (PTQ/QAT) and LoRA/QLoRA.
• Architect model conversion/compilation pipelines using TensorRT for edge deployment.
• Parity, accuracy recovery, and latency benchmarking vs. PyTorch and edge binaries.
• Write and optimize CUDA kernels and TensorRT Plugins for high memory bandwidth and low latency.
• Produce production-grade, concurrent C++ and Python code for real-time edge inference.

🎯 Requirements

• Deep expertise in model quantization (PTQ, QAT) and mixed-precision inference (INT8, FP8, INT4, BF16/FP16).
• Proven experience optimizing large-scale models (LLMs, VLMs, or VLAs) using KV-cache, Speculative Decoding, and Efficient Attention (FlashAttention, Linear Attention).
• Extensive experience with model conversion/compilation pipelines (TensorRT, TensorRT-LLM) and benchmarking.
• Proficiency in low-level programming for AI accelerators: writing and optimizing custom CUDA kernels and TensorRT Plugins.
• Production-level C++ (14/17/20) and Python programming for concurrent, memory-safe, real-time edge inference.

🎁 Benefits

• Zoox Stock Appreciation Rights (SARs) and Amazon RSUs
• Health, long-term care, disability, and life insurance
• Paid time off, vacation, and sick leave
• Competitive compensation and benefits package including health insurance and stock