Founding Research Engineer at Flower

Flower Labs is looking for a Founding Research Engineer to join our new Frontier Model Team. We are a Y Combinator-backed startup known for building the world's most popular open-source framework for decentralized and federated AI training, and we are now applying our unique approach to develop state-of-the-art foundation models that unlock data silos across industries like science, health, and finance.

Responsibilities

• Design and implement training paradigms for frontier models, covering everything from data curation and pre-training to post-training and evaluations.
• Collaborate with a high-impact team to turn conceptual research ideas into functional, scalable systems.
• Conduct systematic experiments to optimize model performance, stability, and scaling behaviors.
• Contribute to the scientific foundation of our models while ensuring they integrate seamlessly into our open-source ecosystem.
• Take on technical leadership responsibilities as our projects grow in complexity and scope.
• Maintain a fast-paced, collaborative workflow where research and engineering expertise are combined to solve challenging problems.

Must-haves

• Deep expertise in current LLM architectures and training methodologies.
• Proven experience with pre-training or post-training processes, such as SFT, RLHF, DPO, or reward modeling.
• Strong grasp of optimization techniques, including mixed precision, stabilization, and scaling laws.
• Proficiency in designing controlled experiments and performing rigorous ablations.
• Fluency in PyTorch or JAX for efficient research prototyping.
• Technical familiarity with Linux, Docker, and version control.
• Excellent written English and a commitment to transparent, honest communication.

Nice-to-haves

• A PhD or Masters degree in a relevant technical field.
• Experience with distributed training frameworks like FSDP, ZeRO, or tensor and pipeline parallelism.
• Background in running large-scale experiments on multi-node or multi-GPU clusters.
• Expertise in safety alignment or advanced preference modeling at scale.
• A strong track record of contributions to open-source projects or high-quality research publications.

Benefits

• Fully remote work environment.