• You want to push the boundaries of small language models capabilities and open-source best-in-class checkpoints
• You actively follow the latest reinforcement learning and optimization research and strive to put theory into practice
• You're equally comfortable crafting domain-specific training environments and profiling GPU utilization in multi-turn asynchronous rollouts

Required Experience

• Strong Python and PyTorch proficiency, with hands-on experience optimizing training pipelines
• Hands-on experience with reinforcement learning and the ability to translate optimization techniques from theory into practical implementations
• Track record of integrating research ideas into robust, maintainable code
• Experience with frameworks like DeepSpeed, FSDP, or vLLM for efficient model training and inference
• Experience working with data pipelines, including curation, validation, and analysis to support post-training objectives
• Contributions to open-source machine learning projects
• M.S. or Ph.D. in Computer Science, Electrical Engineering, Mathematics, or a related field

What You'll Actually Do

• Profile, optimize, and scale RL training runs to reduce iteration time
• Integrate new optimization techniques as they emerge from the research community
• Design and implement tools and environments that test the boundaries of model capabilities
• Turn proof-of-concept ideas into robust training pipelines and best-in-class models

What You'll Gain

• The opportunity to work directly on state-of-the-art AI systems at one of the most advanced AI companies in the world
• A fast-paced, collaborative environment where your work has direct impact on model performance and product capability
• The satisfaction of knowing your craftsmanship helps define the next frontier in AI

About Liquid AI Spun out of MIT CSAIL, we’re a foundation model company headquartered in Boston. Our mission is to build capable and efficient general-purpose AI systems at every scale—from phones and vehicles to enterprise servers and embedded chips. Our models are designed to run where others stall: on CPUs, with low latency, minimal memory, and maximum reliability. We’re already partnering with global enterprises across consumer electronics, automotive, life sciences, and financial services. And we’re just getting started.