Data Partnerships

Generalist•Somerville, MA

About The Position

Generalist is building embodied foundation models for the physical world, beginning with dexterity. The Data Partnerships role owns the data engine that supplies these models. Because robotics data must be collected by people in real environments rather than scraped from the internet, the role finds and scales partners that can collect data to the quality and volume required by model science and scaling laws. Responsibilities span partner sourcing and diligence, hardware support, data specifications, commercial terms, quality oversight, and the network infrastructure for data ingestion and distribution. The work also includes expanding into new geographies and data modalities on tight timelines. A central challenge is translating model needs and data requirements into instructions external partners can execute independently, while identifying when technically compliant data is not practically useful. The role collaborates with ML and engineering leads, establishes operating cadences and clear reporting on cost, volume, and quality, and closes the loop with research by reporting on outcomes and revising subsequent data specifications. Periodic domestic and international travel to partner sites and collection operations is expected.

Requirements

At least 6 years of experience owning external partners or vendors responsible for delivering a product or service, such as in supply chain, vendor management, outsourcing/BPO management, data operations, or partnerships with delivery obligations.
Demonstrated experience standing up a new vendor from selection and diligence through contracting, ramp-up, and ongoing management.
Experience translating technical or specialist requirements into unambiguous instructions that non-technical execution partners can successfully follow.
Fluency with commercial mechanics, including pricing structures, SOWs, acceptance criteria, and unit economics, and judgment about terms that protect quality.
Exceptional written communication, including the ability to write documents that people can execute without the author present.
Comfort operating with incomplete requirements and re-scoping work mid-flight while maintaining the relationship.
Even-keeled under pressure and consistent in follow-through.

Nice To Haves

Experience buying or managing data collection, annotation, or labeling at scale through a managed workforce, crowd, or BPO.
Experience working close to an ML or research organization and sufficient working understanding of training data to challenge requests rather than simply relay them.
Background in autonomous vehicles, robotics, hardware, or another domain where physical-world data acquisition is a first-class problem.
Experience at a rapidly growing company where inherited processes stopped working and had to be rebuilt.

Responsibilities

Develop data supply strategy, deciding what to buy versus build and mapping vendors across data types and quality.
Source, evaluate, and pilot data partners through structured trials before committing volume.
Elicit requirements from ML and engineering leads, including volume, environments, embodiments, task diversity, annotation schema, acceptance criteria, and downstream needs.
Write clear partner-facing specifications defining delivery units, acceptance and rejection criteria, and procedures for ambiguous cases.
Structure pricing, rate cards, minimums, milestones, and acceptance terms; work with legal on MSAs and SOWs.
Own unit economics, measuring cost per unit of usable data rather than merely per unit delivered.
Scale partnerships into a network by bringing on additional partners without sacrificing quality or driving up costs.
Maintain trackers, weekly partner reviews, forecasts against the research roadmap, and accessible reporting on cost, volume, and quality.
Report data outcomes to research and use those results to improve specifications for subsequent collection cycles.
Travel periodically, domestically and internationally, to partner sites and collection operations.
Support data ingestion and distribution infrastructure for the data partnership network and future commercial partnerships.
Extend the data engine into new geographies and data modalities on tight timelines.