Edge AI Research Scientist: Generative Models & Autonomy

Anduril Industries; Costa Mesa (CA); On-site; USD 220,000 - 292,000; Full time

Benefits offered by this job

Top-tier benefits

Equity grants

Job summary

Anduril Industries is seeking an AI Research Scientist to join the Air Dominance & Strike team, designing, fine-tuning, and deploying the next generation of generative AI, LLMs, and agentic systems powering air-dominance platforms and collaborative autonomous behaviors.

This is a highly applied research role focused on making state-of-the-art LLM models smaller, faster, and smarter, with work on both offboard planning and onboard edge hardware.

Qualifications

• Strong Python and DL framework skills (PyTorch/JAX).
• Experience training/evaluating LLMs, Generative AI, or multimodal models.
• 2+ years in CV, NLP, robotics, or related fields focusing on generative models.
• Experience with Axolotl, Hugging Face, DeepSpeed, Megatron-LM, LangChain, or LlamaIndex.
• Eligible to obtain and maintain an active U.S. Top Secret clearance.

Responsibilities

• Develop, pre-train, and fine-tune in-house LLMs and multimodal models.
• Architect and optimize models to run on tactical edge compute and onboard hardware.
• Design and implement robust agentic architectures and multi-agent planning loops.
• Collaborate with CV, perception, and motion planning teams across modalities.
• Define data collection strategies across assets and simulations; build scalable eval frameworks.
• Build early prototypes with customers and scale to production features.

Skills

Python

Deep learning

LLMs training

Transformer models

Edge computing

Education

PhD in CS/ML/Robotics

Tools

Axolotl

Hugging Face

DeepSpeed

Megatron-LM

LangChain

LlamaIndex

Job description

Anduril Industries is seeking an AI Research Scientist to join the Air Dominance & Strike team, designing, fine-tuning, and deploying the next generation of generative AI, LLMs, and agentic systems powering air-dominance platforms and collaborative autonomous behaviors.

This is a highly applied research role focused on making state-of-the-art LLM models smaller, faster, and smarter, with work on both offboard planning and onboard edge hardware.