Video Data & Model Evaluation Specialist

CANTINA RESEARCH SINGAPORE PTE. LTD.; Singapore; On-site; SGD 110,000 - 160,000; Full time

Benefits offered by this job

Health insurance

Global travel insurance

Monthly stipend SGD 500

Equipment for home office

Job summary

Cantina Labs is hiring a Member of Technical Staff to shape data and evaluation strategies for our video generation models in Singapore. You will own data requirements, annotation standards, and high-quality datasets, working with researchers, ML engineers, and external data partners to strengthen data-model feedback loops.

You will build labeling systems, design annotation standards, and create synthetic-data workflows, enabling model iteration and performance attribution across projects.

Qualifications

• Hands-on experience configuring data strategies for multimodal models.
• Experience building high-quality datasets for generative video/image tasks.
• Ability to define annotation standards and evaluation frameworks.
• Experience evaluating model outputs and translating findings into data-driven improvements.
• Familiarity with vision-language models and data automation tools.

Responsibilities

• Define quality standards for video generation across aesthetics, realism, and consistency.
• Own end-to-end data projects including sourcing, annotation, and delivery.
• Develop labeling systems for dataset segmentation and quality filtering.
• Create and evaluate synthetic-data workflows for training value.
• Coordinate with researchers, engineers, product teams, and external data partners.

Skills

Data strategy

Dataset development

Annotation

Model evaluation

Generative media

Communication with partners

Tools

Python

SQL

Vision-language models

Prompt design

Job description

Cantina Labs is hiring a Member of Technical Staff to shape data and evaluation strategies for our video generation models in Singapore. You will own data requirements, annotation standards, and high-quality datasets, working with researchers, ML engineers, and external data partners to strengthen data-model feedback loops.

You will build labeling systems, design annotation standards, and create synthetic-data workflows, enabling model iteration and performance attribution across projects.