Internship 2027 Onboard Infrastructure Engineer, ML Inference

About this role

Bedrock Robotics seeks an intern to integrate large language models and vision-language-action models into autonomous heavy machinery's real-time control stack. You'll optimize multi-billion parameter models for edge deployment on NVIDIA Jetson hardware while maintaining strict latency requirements for construction equipment autonomy.

What you'll do

• Integrate open-source and proprietary LLM/VLA models into Rust-based onboard middleware alongside perception and control pipelines
• Profile and optimize model execution using TensorRT, vLLM, ExecuTorch, and custom edge inference runtimes for NVIDIA Jetson Thor
• Streamline sensor tokenization from cameras and LiDAR to feed real-time streams into models without introducing control loop latency
• Identify and eliminate performance bottlenecks across memory bandwidth, compute, and inter-process communication using profiling tools
• Validate optimizations directly on autonomous heavy machinery at company test sites

What they're looking for

• Rust or C++
• PyTorch or JAX
• GPU architecture and CUDA
• Parallel computing and multithreading
• OS and GPU scheduling, memory management
• Model optimization techniques (quantization, KV-cache management)
• Neural network deployment on edge hardware
• Robotics or multi-modal model experience