Research Engineer

Stellon Labs Inc. is an AI research lab focused on making tiny frontier models for edge devices. Backed by Y Combinator (S25) and other top-tier investors, we're a fast-moving team in San Francisco shipping real research in efficient AI. If you want to do foundational work in model compression, quantization, and efficient inference, we'd love to hear from you..

• Develop novel architectures and training methodologies optimized for edge deployment scenarios
• Design and implement advanced quantization techniques to compress neural networks while maintaining accuracy
• Research and prototype new approaches to model compression, pruning, and knowledge distillation
• Publish findings and contribute to the broader research community
• Mentor junior engineers and drive technical excellence across the team

• Quantization Expertise: Deep understanding of post-training quantization, quantization-aware training, and mixed-precision techniques
• PyTorch Mastery: Extensive experience with PyTorch for research and production, including custom operators and model optimization
• Model Training: Proven track record in training large-scale models, distributed training, and hyperparameter optimization
• Master's or Ph.D. in CS, EE, Physics, or a related field (or equivalent experience) preferred
• Publications in top-tier ML conferences/journals preferred

• Familiarity with training and inference of ultra-low precision models. Experience in writing CPU kernels would make one a strong fit
• Background in writing CUDA(or similar) kernels.
• Experience with ONNX, Tensor
  
  RT, or other model optimization frameworks
• Knowledge of hardware accelerators (GPUs, TPUs, custom silicon)
• Background in computer vision, NLP, or other specific ML domains
• Open source contributions to ML optimization projects