Lead Large-Scale RL & Mid-Training Infra Engineer

Engg; Palo Alto (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Peano AI is seeking a Founding Large-Scale Mid-Training/RL Infrastructure Engineer to design, run, and scale our in-house training stack. You will work with thousands of GPUs and TPU accelerators to optimize pretraining and RL-based methods for large language models, impacting core products and future roadmap.

This role is deeply technical, with responsibilities spanning data pipelines, sharding, checkpointing, and accelerator-aware scheduling.

Qualifications

• Experience with large-scale deep learning model training and distributed systems.
• Proven track record with pretraining and RL-based fine-tuning of large models.
• Hands-on with Megatron, Transformer-Engine, verl, slime.
• Experience in data pipelines, model/data/optimizer sharding, and checkpointing.
• Strong accelerator/memory optimization skills at cluster scale.
• Excellent debugging and performance-tuning in distributed environments.
• Comfort working in deeply technical, early-stage startup settings.

Responsibilities

• Build, optimize, and scale distributed training infrastructure for foundation models.
• Own throughput, efficiency, scalability, cost, and reliability of pretraining and RL pipelines.
• Architect data loading, sharding, checkpointing, batching for multi-node training.
• Integrate and optimize RL components: rollout, reward modeling, environment orchestration.
• Work with Megatron, Transformer-Engine, verl, slime and other libraries.
• Tune memory, mixed-precision, and runtime performance for massive models.
• Debug and profile training bottlenecks across code, compute, networking, and infra.
• Collaborate with research, infra, and applications teams to deliver performant models.

Skills

Large-scale model training

Distributed systems

RL training

Performance tuning

Debugging

Startup mindset

Tools

Megatron

Transformer-Engine

verl

slime

Job description

Peano AI is seeking a Founding Large-Scale Mid-Training/RL Infrastructure Engineer to design, run, and scale our in-house training stack. You will work with thousands of GPUs and TPU accelerators to optimize pretraining and RL-based methods for large language models, impacting core products and future roadmap.

This role is deeply technical, with responsibilities spanning data pipelines, sharding, checkpointing, and accelerator-aware scheduling.