ICLR 2026

📋 Description

• • Research Engineer or Scientist role focused on building foundation models from scratch using Liquid AI's novel hybrid architecture, delivering faster inference and lower memory deployment on edge devices.
• • Day-to-day work involves researching efficient architectures (state space models, hybrid attention, neural ODEs), multimodal vision/audio systems, data engineering for pre-training, infrastructure optimization, or post-training alignment techniques, with end-to-end ownership from idea to shipped model.
• • Small, high-impact team spun out of MIT CSAIL that publishes research, releases open weights, and presents at top conferences like ICLR, NeurIPS, and CVPR.
• • Opportunity to own entire workstreams, contribute to visible research that ships, and advance the science of efficient multimodal foundation models in a unicorn-stage AI startup.