Data Scientist (AI Data & LLM Specialist)

Stack

Responsibilities

• ▹Develop data labeling strategies: design and document a formal data annotation strategy with clear, scalable and efficient labeling guidelines
• ▹Define and enforce quality metrics, including inter-annotator agreement
• ▹Optimize for LLM consumption: research, define and prototype the data formats, structures and pre-processing steps needed for fine-tuning and training LLMs on the datasets
• ▹Data quality analysis: establish automated processes and metrics to analyze the quality of raw and labeled data, feeding back into data collection and labeling workflows
• ▹Collaborate with engineering to guide the implementation of data processing pipelines and ensure the data infrastructure meets the needs of ML applications

Requirements

• ▹Proven experience as a Data Scientist or Machine Learning Engineer with a focus on data quality and preparation
• ▹Strong understanding of data labeling methodologies and hands-on experience with data annotation platforms and workflows
• ▹Experience preparing datasets for training and fine-tuning LLMs, including tokenization, embeddings and NER
• ▹Proficiency in Python and common data science libraries (e.g. Pandas, NumPy, Scikit-learn, spaCy, Hugging Face)
• ▹Experience using APIs/SDKs to automate data annotation and active learning loops
• ▹Excellent communication skills, with the ability to create clear documentation for technical and non-technical audiences

Nice to have

• ▹Experience with audio data processing and relevant libraries
• ▹Familiarity with data annotation platforms and tools
• ▹Knowledge of modern MLOps principles and practices
• ▹Experience with LLM data curation and Reinforcement Learning from Human Feedback (RLHF) pipelines

Soft skills

What we offer

• ▹Competitive salary, equity and benefits package
• ▹Synchronous and asynchronous collaboration: weekly all-hands meetings, Slack, quarterly in-person meetups

About the company

Eclipse is building the fastest Ethereum Layer 2, powered by the Solana VM, and develops its own apps on top. It is backed by investors including Polychain, Tribe Capital, Placeholder and DBA, and its founding team has experience launching projects such as dYdX, Uniswap and zkSync.