Teams at Cantina Labs

Conduct research on next-generation video generation models, focusing on distillation, inference efficiency, reward modeling, preference optimization, evaluation, and scalable training infrastructure. Interns will formulate hypotheses, run large-scale experiments, analyze results, develop research tooling, and communicate findings through reports, presentations, demonstrations, and potential conference submissions under senior mentorship.

Three-month onsite internship developing and evaluating video-generation machine learning systems. Projects may involve reward modeling, inference efficiency, multimodal data pipelines, distributed training infrastructure, large-scale experimentation, and evaluation tooling. Interns own a defined research or engineering project with senior mentorship and may contribute to conference publications or product-adjacent work.

Build and productionize large-scale TTS and adjacent speech models end-to-end: model architecture, training, alignment, evaluation, data curation, distributed GPU scaling, latency/cost profiling, tooling, and safety/mitigation for speech systems.