About the role
Nouvells is seeking AI/ML Research Engineers who combine strong technical engineering skills with curiosity, experimentation, and research-oriented thinking. You'll investigate emerging approaches, develop prototypes, evaluate models, and translate promising research into practical AI systems.
As our AI/ML Research Engineer, you will drive core algorithmic breakthroughs and architecture development for Nouvell. You will read the latest preprint on Monday, evaluate its mathematical viability on Tuesday, and have an instrumented, distributed prototype training on our GPU clusters by Thursday. You will focus on fundamental model capabilities: designing novel Transformer and hybrid architectures (e.g., Mixture of Experts, State Space Models), advancing post training reasoning paradigms (Reinforcement Learning with Verifiable Rewards, multi step chain of thought distillation), authoring custom GPU kernels, and pushing the pareto frontier of model capability versus inference compute.
Key responsibilities
- Design, implement, and benchmark modern neural network architectures
- Formulate and execute large-scale pre-training and continual pre-training runs across multi-node GPU clusters, managing loss spikes, gradient stability, and data curriculum scheduling.
- Explore test-time compute paradigms, tree-search decoders, and process-supervised reward models (PRMs) to enhance inference-time problem-solving fidelity.
- Rapidly digest, reproduce, and critically evaluate frontier papers from venues such as NeurIPS, ICML, ICLR, and arXiv.
- Maintain rigorous experimental reproducibility, tracking ablation studies, training dynamics, and model checkpoints using distributed telemetry tools.
Qualifications
- Bachelor’s degree in Computer Science and any related field