Research Engineer

Goliath Partners Inc.; New York (NY); On-site; USD 120,000 - 180,000; Full time

Job summary

Goliath Partners Inc. is seeking a Research Engineer (Datasets) to advance multimodal world models. You will own the data research lifecycle, from designing datasets and running experiments to building scalable synthetic data pipelines and evaluating model performance.

Join a fast-growing AI team that values curiosity, rigorous experiments, and responsible data practices. You’ll collaborate with research, product, and creative teams to translate goals into data strategy and drive meaningful

Qualifications

• 4+ years of experience in ML or AI.
• Experience with large multimodal datasets and generative models.
• Understanding of how data quality impacts model performance.
• Experience across the ML lifecycle: dataset creation, model training, evaluation, and experimentation.
• Proficiency with PyTorch, JAX, or similar frameworks.
• Experience with distributed computing tools such as Ray or Kubernetes.
• Curious, strong research instincts, and an experimental mindset.

Responsibilities

• Design multimodal datasets for foundational models.
• Run controlled experiments to study data composition effects on performance.
• Build and scale synthetic data generation, filtering, and QC pipelines.
• Develop evaluation frameworks and benchmarks to measure improvements.
• Partner with research, product, and creative teams to translate goals into data strategies.
• Iterate on datasets and training pipelines to improve world model capabilities.

Skills

4+ years of ML experience

Multimodal datasets experience

Data quality & composition

ML lifecycle expertise

PyTorch / JAX

Distributed computing (Ray / K8s)

Research mindset / curiosity

Tools

PyTorch

JAX

Ray

Kubernetes

Job description

A fast-growing, venture-backed AI company is hiring a Research Engineer (Datasets) to help build the next generation of multimodal world models. This role sits at the intersection of machine learning research and data engineering, focusing on how high-quality datasets unlock new model capabilities. You'll own the full data research lifecycle, from designing datasets and running experiments to building scalable synthetic data pipelines and evaluating model performance.

What You'll Do:

• Design multimodal datasets that teach foundation models new capabilities
• Run controlled experiments to understand how data composition impacts model performance
• Build and scale synthetic data generation, filtering, and quality control pipelines
• Develop evaluation frameworks and benchmarks that measure meaningful model improvements
• Partner closely with research, product, and creative teams to translate product goals into data strategies
• Iterate on datasets and training pipelines to continuously improve world model capabilities

What You Bring:

• 4+ years of experience in machine learning or applied AI
• Experience working with large multimodal datasets and generative AI models (image, video, or multimodal)
• Strong understanding of how data quality and composition influence model performance
• Experience across the ML lifecycle, including dataset creation, model training, evaluation, and experimentation
• Proficiency with PyTorch, JAX, or similar ML frameworks
• Experience with distributed computing tools such as Ray, Kubernetes, or equivalent infrastructure
• Curiosity, strong research instincts, and an experimental mindset

Why Join:

• Work on cutting-edge multimodal foundation models and world model research
• Own one of the most impactful parts of the AI stack, data strategy and model capability development
• Collaborate with a small, world-class team pushing the frontier of generative AI
• High ownership with the ability to influence research direction and product capabilities
• Backed by leading investors and building technology at the forefront of multimodal AI

If you're excited about building the datasets that power the next generation of world models and want to work on some of the most challenging problems in AI, I\'d love to connect.