Member of Technical Staff - Engineering Lead, Data Ingestion

Data Ingestion Lead

Reflection's Data team builds the training corpora our frontier models learn from. Before a model can learn anything, the data has to be found, fetched, extracted, and delivered reliably, responsibly, and at enormous scale. The ingestion layer is the machinery that turns the open web, licensed corpora, and other large-scale sources into well-structured, versioned, auditable datasets for pre-training.

As Data Ingestion Lead, you'll provide front-line leadership of the team that builds this layer, spanning all three of its pillars web crawl, data ingestion pipelines, and data lakes. You'll build, mentor, and grow a team of data ingestion engineers, guide the technical and architectural decisions across crawling, extraction, and corpus storage/delivery, and work closely with the pre-training research, data quality, and data partnerships teams that depend on what you ship.

You'll stay close enough to the stack to make targeted contributions as an individual contributor and to maintain a deep understanding of the team's technical work. Subtle decisions at the ingestion layer what we crawl, how we extract, what we keep ripple through training and directly affect where our models are strong, safe, and where they fail.

• Build, mentor, and grow a high-performing team of data ingestion engineers. Coach and support your reports in understanding, and pursuing, their professional growth.

• Provide front-line leadership across the full ingestion stack — web crawling and acquisition, extraction and normalization pipelines, and the data lakes that version and deliver training corpora and the campaigns that run across all three.

• Stay hands-on: become familiar with the team's technical stack enough to make targeted contributions as an individual contributor.

• Manage day-to-day execution: prioritize the team's work and run data acquisition and ingestion campaigns in a highly dynamic, fast-paced environment.

• Guide technical and architectural decisions, emphasizing scalability, reliability, auditability, and cost crawler architecture and politeness/scheduling, distributed processing (Ray/Beam/Spark), orchestration (Airflow/Prefect), storage formats and layout (Parquet, JSONL, WARC; object stores and lakehouse formats), deduplication, and dataset versioning and delivery.

• Close the loop with research: partner with pre-training and data quality teams to tie ingestion decisions to measurable downstream model impact, and run experiments to evaluate crawling strategies, extraction methods, and ingestion tradeoffs.

• Work across the wider data effort data partnerships, digitization operations, and external vendors to onboard new sources within legal, licensing, and robots.txt constraints.

• Raise the bar for technical judgment, prioritization, communication, and execution in a fast-moving environment.

• Experience building, mentoring, and growing data or infrastructure engineering teams while staying technically hands-on. (Comfortable growing into leading a larger team quickly if you haven't managed at that scale before.)

• Deep experience building web-scale data acquisition or ingestion systems, with real ownership of production-grade pipelines at multi-TB to PB scale.

• Strong coding ability and the credibility to earn the technical trust of a strong team.

• Deep expertise in at least one of web crawling & acquisition, large-scale extraction & ingestion pipelines, or data lakes / corpus storage & delivery with working knowledge across the others, and the ability to learn the rest.

• Fluency with the modern large-scale data toolkit: distributed compute (Ray, Beam, Spark), orchestration (Airflow, Prefect), formats (Parquet, JSONL, WARC), and object-store / data-lake architectures.

• Familiarity with how LLMs are trained and evaluated, and an intuition for what makes data useful for training; comfortable designing experiments and using proxy quality signals to guide system improvements.

• Ability to guide strategy and drive execution across concurrent data campaigns, and to partner effectively with research teams, external data vendors, and partnerships/legal on acquisition.

We believe…