Research Scientist, Data-Centric AI for Faster Models

datologyai; San Mateo; On-site; PHP 11,264,000 - 18,773,000; Full time

Benefits offered by this job

Health benefits

401(k) plan with company match

Unlimited PTO

Parental leave

Wellness stipend

Learning stipend

Daily lunches

Relocation assistance

Job summary

DatologyAI is seeking a Research Scientist to explore how intervening on training data can shape deep learning models. You will source ideas from literature, ground research in customer needs, and collaborate with engineers and product teams to convert findings into real products.

You will work autonomously in a fast-moving startup and contribute to data-driven improvements across vision, language, or multimodal models, with a strong emphasis on reproducibility and scalability.

Qualifications

• 3+ years of deep learning research experience.
• Strong fundamentals in deep learning.
• Experience or publications in data pruning and curation.
• Experience with curriculum learning, synthetic data generation, dataset distillation.
• Awareness of how training data affects model behavior and embedding models with semantic search.
• Experience training large vision, language, or multimodal models.
• Knowledge of efficient ML concepts.
• Proficiency in PyTorch and software engineering practices.

Responsibilities

• Source, vet, implement, and improve ideas from the literature and your own thinking.
• Collaborate with engineers and product teams to turn findings into tangible impact.

Skills

Deep learning

PyTorch

Software engineering

Communication & collaboration

Research publications

Tools

PyTorch

Spark

Snowflake

Job description

DatologyAI is seeking a Research Scientist to explore how intervening on training data can shape deep learning models. You will source ideas from literature, ground research in customer needs, and collaborate with engineers and product teams to convert findings into real products.

You will work autonomously in a fast-moving startup and contribute to data-driven improvements across vision, language, or multimodal models, with a strong emphasis on reproducibility and scalability.