Research Scientist for real-time video generation

Cantina Labs; United States; A distancia; USD 150.000 - 230.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Medical, dental, vision

42 days PTO

Parental leave

Fertility support

401(k)

Lifestyle allowance

Lunch & snacks

One Medical

Company equity

Descripción de la vacante

Cantina Labs ищет инженера по разработке и улучшению видеогенеративных моделей в реальном времени. Ваша работа будет направлена на создание интерактивных сценариев, поддержание качества и согласованности персонажей во время длительной генерации.

Требуется опыт обучения видеогенеративных моделей, способность быстро осваивать новые методы и проводить надежные эксперименты. В приоритете ясная коммуникация и ответственность, работа в кросс-функциональной команде.

Formación

• Опыт обучения или донастройки видеогенеративных моделей.
• Умение быстро осваивать новые методы и превращать идеи в полезные эксперименты.
• Хорошие показатели экспериментального дизайна, ответственность, ясная коммуникация.

Responsabilidades

• Разработка моделей генерации видео в реальном времени и потоковую обработку с сохранением качества и согласованности.
• Изучение задержек, непрерывности и изменений поведения моделей при длительной генерации.
• Исследование взаимодействия внешности, мимики, жестов и движений персонажей.

Conocimientos

Video generative models

Experimentation skills

Communication & ownership

Descripción del empleo

Описание:

Cantina Labs is a social AI company developing advanced real-time models for expression, personality, and realism. Its flagship product, Cantina, is a social AI platform that enables people to tell stories, connect, and create.

Задачи:

• Develop real-time and streaming video generation models that support responsive interaction while maintaining visual quality and consistency
• Investigate latency, continuity, and changes in model behavior during longer generation
• Explore how appearance, facial expression, gestures, and body movement combine into coherent characters
• Improve the relationship between speech and motion across identities and visual styles
• Adapt video models to new data, behaviors, and capabilities
• Investigate training objectives and conditioning methods, and implement and evaluate promising ideas
• Improve generation speed and efficiency through model-level approaches and runtime improvements
• Measure the effects of optimization on quality, latency, and memory
• Identify gaps in training data and recurring failure modes; improve curation and sampling, and develop data to teach desired behaviors
• Build evaluations that make progress visible and failures understandable, combining quantitative measurements with inspection of generated video and interaction experience
• Read relevant research, reproduce and adapt useful ideas, and share early experiments, demos, and findings
• Help the team decide which ideas merit more time and compute

Требования:

• Experience training or fine-tuning video or related generative models
• Ability to learn unfamiliar methods quickly and turn ideas into useful experiments
• Good experimental judgment, ownership, and clear communication

Nice to have:

Условия:

• Competitive salary and generous company equity
• For U.S.-based roles: medical, dental, and vision insurance, with 99.99% of premiums covered by Cantina
• For U.S.-based roles: 42 days of paid time off, including 15 PTO days, 10 sick days, 15 company holidays, and 2 floating holidays
• For U.S.-based roles: generous parental leave and fertility support
• For U.S.-based roles: 401(k) retirement savings plan
• For U.S.-based roles: lifestyle spending account of $500/month
• For U.S.-based in-office employees: complimentary lunch and snacks
• For U.S.-based roles: One Medical membership