Hybrid Senior Research Scientist, LLMs

at Aldea

• Job title: Senior Research Scientist, LLMs
• Job type: Full Time
• Experience level: Senior
• Salary: Not specified
• Degree requirement: Postgraduate Degree
• Tech skills: Node.js, PyTorch
• Location requirements: Hybrid, San Francisco, United States

About the role

• Foundational AI Research Scientist developing next-generation language models. Pioneering large-language-model architectures and attention mechanisms for efficient scaling.

Responsibilities

• Research and prototype sub-quadratic attention architectures to unlock efficient scaling of large language models.
• Design and evaluate efficient attention mechanisms including state-space models (e.g., Mamba), linear attention variants, and sparse attention patterns.
• Lead pre-training initiatives across a range of model scales from 1B to 100B+ parameters.
• Conduct rigorous experiments measuring the efficiency, performance, and scaling characteristics of novel architectures.
• Collaborate closely with product and engineering teams to integrate models into production systems.
• Stay at the forefront of foundational research and help shape Aldea's long-term model roadmap.

Requirements

• Requires a Ph.D. in Computer Science, Engineering, or related field.
• 3+ years of relevant industry experience.
• Deep understanding of modern sequence modeling architectures including State Space Models (SSMs), Sparse Attention mechanisms, Mixture of Experts (MoE), and Linear Attention variants.
• Hands-on experience pre-training large language models across a range of scales (1B+ parameters).
• Expertise in PyTorch, Transformers, and large-scale deep-learning frameworks.
• Proven ability to design and evaluate complex research experiments.
• Demonstrated research impact through patents, deployed systems, or core-model contributions.
• Nice to Have Experience with distributed training frameworks and multi-node optimization.
• Knowledge of GPU acceleration, CUDA kernels, or Triton optimization.
• Publication record in top-tier ML venues (NeurIPS, ICML, ICLR) focused on architecture research.
• Experience with model scaling laws and efficiency-performance tradeoffs.
• Background in hybrid architectures combining attention with alternative sequence modeling approaches.
• Familiarity with training stability techniques for large-scale pre-training runs.

Benefits

• Competitive base salary
• Performance-based bonus aligned with research and model milestones
• Equity participation
• Flexible Paid Time Off
• Comprehensive health, dental, and vision coverage