Research Scientist / Engineer – Foundation Model: Core Research

lumalabs-ai; San Francisco (CA); On-site; USD 250,000 - 450,000; Full time

Job summary

lumalabs-ai is seeking a senior researcher to shape the future of multimodal AI, bridging research with shipped products like Dream Machine and Ray3. You will work at the intersection of modeling, data, systems, and evaluation to advance state-of-the-art in multimodal intelligence.

Responsibilities include designing scalable architectures, evaluating alignment, and building research infrastructure for fast experiments on large-scale GPU clusters.

Qualifications

• Degree in CS/ML/Physics/Math (any of these) is essential.
• First-principles intuition for scaling and architecture behavior at scale.
• Fluency in frontier AI language, research and engineering as a unified discipline.
• Ability to design and analyze experiments and communicate complex concepts.

Responsibilities

• Drive core research that powers all products, co-designing multimodal representations and scaling laws.
• Develop proxy tasks and automated metrics to align training with user experience beyond benchmarks.
• Build production-research parity and infrastructure to accelerate rapid experimentation.

Skills

First-principles mindset

Experiment design & analysis

Multimodal AI knowledge

Distributed computing

GPU cluster management

Education

Bachelor's/Master's/PhD in CS/ML/Physics/Math

Tools

GPU clusters

Job description

Where You Come In

This is a rare and foundational opportunity to define the future of multimodal AI. You will be at the forefront of architecting the intelligence governing our world-simulations—the reasoning core at the heart of our world-modeling efforts. This role offers the chance to bridge frontier research with magical, shipped products like Dream Machine and Ray3, solving novel problems where no playbook exists.

What You'll Do

This opportunity involves both the “science” and “engineering” of research—two aspects we view as equally critical. You will work at the intersection of modeling, data, systems, and evaluation to advance the state-of-the-art in multimodal intelligence.

• Unified Modeling & Efficiency Drive the core research that powers all of Luma's products — co-designing multimodal representations, advancing core algorithms for long-context training, and establishing rigorous scaling laws to predict performance across compute budgets.


• Alignment & Evaluation Close the gap between training loss and user experience. Develop proxy tasks and automated metrics that serve as the compass for research decisions — ensuring our models optimize for what actually matters to users, not just benchmarks.


• Research Infrastructure Build the engine for high-velocity research. Maintain production-research parity, ensure reproducibility, and design systems for rapid experimentation — so that novel ideas go from hypothesis to validated result as fast as possible.

Who You Are

• A Bachelor's, Master's, or PhD degree in Computer Science, Machine Learning, Physics, or Mathematics is essential.


• A 'first-principles' intuition for scaling. You don't just follow the literature; you understand why certain architectures succeed or fail at scale.


• Fluent in the language of frontier AI. You see research and engineering as a single, unified discipline.


• Proven ability to design and rigorously analyze experiments and to articulate complex technical concepts effectively.


• Practical experience with distributed or high-performance computing environments, particularly managing and optimizing training runs on large-scale GPU clusters.

What Sets You Apart (Bonus Points)

• A track record of publishing at top-tier venues (NeurIPS, ICML, ICLR) and a mission-driven, \"first-principles\" mindset.


• Infrastructure Expertise: Proven ability to build and lead research infrastructure for technical teams, ensuring production-research parity.


• Engineering Excellence: Strong commitment to software engineering best practices, including optimizing for code readability and reusability, implementing comprehensive unit and integration tests, and maintaining high documentation standards (necessary docstrings).


• Experience with low-precision training and hardware-aware optimization for next-gen clusters.

Your application are reviewed by real people.

Compensation

The base pay range for this role is $250,000 – $450,000 per year.