AI Research Scientist

Wirestock; San Francisco (CA); On-site; USD 150,000 - 220,000; Full time

Job summary

Wirestock is seeking an AI Research Scientist to architect the data lab that powers multimodal models. You will own the data frontier, analyzing latest models to design datasets that reveal gaps and enable superior training.

You will build ground-truth architectures, guide data collection across real and synthetic sources, and collaborate with leading AI researchers to validate concepts at scale.

Qualifications

• Master’s or Ph.D. in CS/ML or equivalent research experience.
• Experience discussing dataset biases and evaluation protocols.
• Strong ability to design high-quality multimodal datasets.

Responsibilities

• Own the data frontier by designing datasets that address model gaps.
• Architect ground truth for multimodal systems to improve training.
• Serve as technical liaison with top AI labs to advance projects.
• Run data experiments and prototypes to validate hypotheses at scale.

Skills

Dataset bias measurement

Synthetic vs real data mixing

Evaluation protocols

Education

Master’s or Ph.D. in CS/ML

Job description

Wirestock is one of the leading data platforms for ethically sourced multimodal data. We serve some of the world’s top AI labs, including several foundation models, by providing high-quality, fully licensed training datasets. As the AI data landscape undergoes a major shift, we are scaling rapidly to meet rising demand for curated visual data.

Role Description

At Wirestock, we believe that the next leap in AI performance won't come from larger models, but from better data. We are looking for an AI Research Scientist who is obsessed with the Data half of Algorithms. Your role is to be the technical architect of our data lab, identifying what data the world's leading AI labs need before they even realize they need it.

What You Will Own

The Research Frontier: You will be our SOTA scout. You’ll spend your time dissecting the latest models and papers to identify gaps and then design the dataset that fixes it.

Dataset Architecture: You will design the ground truth for multi-modal systems. This means architecting how we use real world data to create datasets that don't just look good, but are mathematically superior for training.

Technical Diplomacy: You will be the peer-level contact for researchers at top-tier AI labs, helping them develop the next big models.

Validation & Prototyping: You will run data experiments using synthetic tools and our global creator network, you’ll build small-scale proof of concept datasets to validate your hypotheses before we scale them to millions of assets.

What You Will NOT Do

Model Training: You are here to build the engine’s fuel, not the engine itself. We leave the model training to our clients so that you can focus 100% of your intellectual energy on data innovation.

Who You Are

The Multimodal Expert: You have a Master’s or Ph.D. in CS/ML or equivalent industry research experience and can talk at length about:

• Dataset bias measurement
• Synthetic vs real data mixing
• Evaluation protocols

Data-Obsessed: You believe that a perfectly curated 10k-video dataset is worth more than a billion-scale "scrape" of the internet.

The Creative Scientist: You are as comfortable discussing the physics of light and motion as you are writing PyTorch scripts to validate CLIP-alignment.