Remote Member of Engineering (Pre-training / Data Acquisition)
Role details
Job location
Tech stack
Job description
You'll be working alongside our pre-training data team, focused on one of the most foundational challenges in training frontier LLMs: acquiring the best possible pre-training data.
The data we collect is upstream of everything. It directly shapes the capability of the models we train. As our first dedicated data acquisition engineer, you will spearhead and evolve systems that crawl the web at massive scale, rapidly ingest data from strategic partnerships, and build specialized tooling to maximize recall from high-value sources. You'll collaborate closely with pre-training data researchers and engineers to ensure that our sourcing of data maps to our training needs, to ensure we have the most capable pre-trained models., * Design, build, and operate a large-scale web crawler responsible for acquiring all openly accessible data on the internet
- Develop specialized deep crawlers targeting high-value sources to improve recall and coverage
- In collaboration with data researchers, own a long-term road map for data acquisition
- Build observability, monitoring, and debugging tooling to ensure reliability and transparency across crawl infrastructure
- Collaborate with pre-training, post-training, and evaluations teams to align data acquisition priorities with model training needs
- Build high-throughput ingestion pipelines for rapidly onboarding partner data and evaluating it for quality
Requirements
- Strong distributed systems background with proven experience building and operating large-scale infrastructure - data pipelines, web crawlers, or similar
- Proficiency in Python, and comfortable optimizing performance and debugging complex systems under production conditions
- Hands-on experience with web crawling or large-scale data extraction: understanding of HTTP protocols, distributed job queues, and data parsing at scale
- Familiarity with cloud platforms (AWS) and container orchestration (Kubernetes, Docker) for deploying and managing high-throughput workloads
- Awareness of the non-technical dimensions of internet-scale crawling: data privacy, robots.txt adherence, and responsible crawl practices
- Nice to have:
- Prior experience pre-training LLMs
- Experience in building trillion-scale SOTA pre-training datasets
- Experience translating research to production at scale
PROCESS
Benefits & conditions
- Fully remote work & flexible hours
- 37 days/year of vacation & holidays
- 16 weeks of flexible, full-pay parental leave
- Health insurance allowance for you & dependents
- Company-provided equipment
- Well-being, always-be-learning & home office allowances
- Frequent team get togethers
- Diverse & inclusive people-first culture