Training Infrastructure Engineer - Remote

Posted 2 weeks ago

Software Development

Full Time

USA

Multimodal Data Loading

Checkpointing Mechanisms

GPU Clusters

Data Pipelines

Performance Optimization

Overview

As a Training Infrastructure Engineer at Liquid, you will design and implement scalable training infrastructure to support the development of advanced AI models, focusing on optimizing performance and efficiency.

In Short

Build distributed training infrastructure for language and multimodal models.
Optimize communication patterns and resolve performance bottlenecks.
Work with diverse data types and implement robust data pipelines.
Design checkpointing systems for efficient training recovery.
Collaborate with ML engineers to enhance model architectures.

Requirements

Experience with frameworks like PyTorch Distributed and DeepSpeed.
Understanding of hardware accelerators and networking.
Ability to identify and resolve training pipeline bottlenecks.
Experience with multimodal datasets and preprocessing.
Familiarity with custom sharding techniques for GPU clusters.

Benefits

Opportunity to tackle complex AI systems challenges.
Experience in building infrastructure for next-gen foundation models.
Direct impact on breakthroughs in AI model capabilities.

Liquid AI

Liquid AI is a foundation model company spun out of MIT CSAIL, headquartered in Boston. The company is dedicated to building capable and efficient general-purpose AI systems that operate at every scale, from mobile devices to enterprise servers and embedded chips. Liquid AI's innovative models are designed to function where others struggle, utilizing CPUs with low latency, minimal memory, and maximum reliability. The company collaborates with global enterprises across various sectors, including consumer electronics, automotive, life sciences, and financial services, and is committed to redefining the architecture of intelligence through advanced AI technologies.

Share This Job!

Save This Job!