We’re looking for an AI Researcher focused on AI architecture research to help design, analyze, and advance next-generation model architectures. You’ll work at the intersection of theory and production—publishing novel research while collaborating closely with engineers to turn ideas into real systems.
This role is ideal for someone who has published research papers and wants to see their work directly shape deployed models, not just benchmarks.
Research and design novel AI architectures (e.g. alternatives to standard Transformer designs, long-context models, efficient sequence modeling, hybrid architectures)
Explore architectural improvements for scalability, efficiency, and stability
Prototype and evaluate new architectures through ablations, benchmarks, and empirical studies
Author and co-author research papers for top ML conferences and journals
Collaborate with engineering teams to translate research into training and inference systems
Stay current with state-of-the-art research and identify promising directions early
Strong background in machine learning research, with a focus on model architecture
Publication record in ML/AI venues (e.g. NeurIPS, ICML, ICLR, COLM, ACL, EMNLP, arXiv)
Deep understanding of:
Neural network architectures
Sequence models and attention mechanisms
Training dynamics and optimization
Hands-on experience with PyTorch or JAX
Ability to reason rigorously, design clean experiments, and communicate results clearly
Comfortable working in a fast-moving startup environment
Experience with non-Transformer architectures (e.g. RNN-based, state-space, hybrid models)
Work on long-context or memory-efficient models
Open-source research contributions
Experience bridging research and production systems
Background in efficient training or inference-aware architecture design
High ownership over research direction and roadmap
Clear path to publishing impactful work
Tight feedback loop between research and real-world deployment
Small, highly technical team with strong research culture
Competitive compensation and meaningful equity
We enable serverless inference via our GPU orchestration and model load-balancing system. We unlock fine-tuning by enabling organizations to size their server fleet to throughput needs, not number of models in the catalogue.
See it in action on our public cloud, which offers inference for 10k+ open weight models.
