Machine Learning Infrastructure Engineer Intern – Machine Learning Infrastructure

Job Description

Machine Learning Infrastructure Engineer Intern | PlusAI

The Tone:
PlusAI is a Physical AI company building AI-based virtual driver software for autonomous trucks. This role focuses on accelerating BEV model training by identifying bottlenecks and implementing high-performance solutions. You will contribute to deploying next-generation autonomous trucks.

The TL;DR
• Role: Machine Learning Infrastructure Engineer Intern
• Mission: Identify and resolve bottlenecks in BEV model training to accelerate the process, exploring LLMs for assistance.
• Tech Stack: CUDA, Triton, C++, PyTorch, Nsight, PyTorch profiler, LLMs

What You’ll Actually Do
• Identify: Profile and analyze Bird’s Eye View (BEV) model training pipelines to pinpoint computational and memory bottlenecks.
• Develop: Design and implement high-performance custom compute kernels using CUDA, Triton, or C++ to accelerate training.
• Leverage: Explore and integrate Large Language Models (LLMs) to assist in generating high-performance code and optimizing kernel logic.
• Automate: Build systems to automate performance profiling and analysis using NVIDIA Nsight and the PyTorch Profiler.

The Must-Haves
• Background: Solid understanding of memory management, computer architecture, and parallel processing principles.
• Experience: Hands-on experience with PyTorch, specifically custom operations, autograd, and training loops. Practical experience writing and optimizing custom GPU kernels using CUDA or Triton is preferred.
• Skills: Strong proficiency in C++, performance tuning, algorithmic efficiency, and low-level system optimization.