Staff AI Training Platform Engineer

Unconventional AI; Los Angeles, California (CA, MO); Sur place; USD 180 000 - 260 000; Plein temps

Avantages offerts par ce poste

Competitive salary and equity

Best-in-class health benefits

401k matching

Unlimited PTO

Complimentary meals in Palo Alto

Résumé du poste

Unconventional AI is building a scalable AI training platform to accelerate model development. You will architect and scale end-to-end training infrastructure, enabling researchers to map neural networks to novel hardware while advancing energy-aware compute.

The role emphasizes deep expertise in distributed training, low-level kernel work with CUDA/Triton, and collaboration across theory and hardware teams to push the envelope of performance and efficiency.

Qualifications

• 5+ years of AI/ML engineering experience.
• Experience mapping state-of-the-art AI models to system performance.
• Strong knowledge of distributed training and parallelism.

Responsabilités

• Architect, scale, and maintain the core AI/ML training platform and infrastructure.
• Develop and optimize training frameworks and pipelines.
• Benchmark MFU, memory bandwidth, and convergence stability across multi-node clusters.
• Collaborate with theory and hardware teams to translate model requirements into infra specs.
• Design and implement robust model checkpointing and recovery mechanisms.

Connaissances

AI/ML engineering

Distributed training

Python

C++

Megatron-LM/DeepSpeed

Formation

BS in Computer Science, Physics, Electrical Engineering, or Applied Math

Outils

Megatron-LM

DeepSpeed

Ray

PyTorch Lightning

Description du poste

Unconventional AI is building a scalable AI training platform to accelerate model development. You will architect and scale end-to-end training infrastructure, enabling researchers to map neural networks to novel hardware while advancing energy-aware compute.

The role emphasizes deep expertise in distributed training, low-level kernel work with CUDA/Triton, and collaboration across theory and hardware teams to push the envelope of performance and efficiency.