Strategic Projects Lead

Foundry Robotics•San Francisco, CA

8h•Onsite

About The Position

Foundry Robotics is building an AI-native robotics manufacturing company focused on deploying advanced assembly and production capability for leading robotics companies and national-security-critical hardware. We are reimagining manufacturing through advanced robotics. Our mission is to rebuild the American manufacturing industry as an AI-first, assembly-focused, dual-use contract manufacturer. We aim to empower manufacturers with intelligent, efficient, and adaptable robotic systems that redefine productivity and quality. This role is for a Strategic Project Lead who thrives on turning model team requirements into production data, equally comfortable architecting QA systems from first principles and calibrating a rig on the floor. You will define what gets captured, how it lands, and whether it is usable, from early task spec through pre-delivery validation. You will stand up the data collection program from scratch and drive rapid build-test-learn cycles to deliver a dataset that is both rigorous and scalable. You will be the day-to-day interface with our model team and the frontier labs we deliver to, and the source of truth for hours collected, hours delivered, and hours usable. You will partner closely with the Operations Manager, Robotics Data, who owns the workforce, floor execution, and daily fleet readiness.

Requirements

3–5 years as a strong PM or TPM with exposure to hardware, data pipelines, or physical workflows
Proven ability to build data collection or technical operations programs from the ground up and improve on existing ones
Comfortable owning task specs, QA programs, and standardization
Equally effective at a desk and on the floor, with deep hands-on experience calibrating a rig one hour and writing a stakeholder update the next
SQL or Python proficient enough to pull your own data and build your own dashboards
Track record tracking and improving program metrics (usable hours, mix accuracy, SLA attainment, rejection rates)
Experience providing technical input in a cross-functional environment with researchers and engineers
Collaborative by nature, you consider downstream impacts on operators, model teams, and infrastructure when making program decisions
Bias toward action, scrappy problem solving, and moving fast without sacrificing rigor

Nice To Haves

Experience supporting vendor or hardware supplier relationships
Exposure to robotics, VLA models, or foundation model training data
Experience scaling a data collection program from prototype to high-volume production
Familiarity with computer vision, sensor calibration, or robotics data pipelines

Responsibilities

Manage the rig fleet end-to-end (UMI grippers, GoPros, VR headsets and controllers), including procurement, lifecycle, asset tracking, spares, and vendor relationships
Own rig calibration cadence and drift detection, including GoPro intrinsics, fisheye calibration, UMI gripper width drift, and per-operator VR eye-tracking calibration
Define success criteria with the model team, translating high-level data needs into rejection rules an assembler can apply consistently
Author SOPs and hand them to the Operations Manager for floor execution
Own new task ramp-up, including spec, SOP, pilot batch design, and quality gate sign-off before full production
Curate gold-standard demos (20–50 reference demos per task), maintained against drift
Run post-capture QA (sampling + success classifier) and pre-delivery validation against model team spec
Track diversity and coverage across task variants, object positions, lighting conditions, and operator styles
Run daily capture logs and weekly dataset reporting to model and research stakeholders, including hours captured, hours usable, mix vs. target, top rejection reasons, and blockers
Own task spec versioning, coordinating SOP updates, retraining triggers, and data v1/v2 marking when the model team changes a spec
Own the 48-hour handoff SLA, including staging, upload integrity, episode schema enforcement, and metadata completeness (operator ID, task ID, success label, timestamp written atomically)
Run RCA when batches get rejected and close the loop with the floor
Define and report against program-level OKRs for usable hours delivered, mix accuracy, and SLA attainment
Own data governance and compliance, including episode-level metadata for auditability, retention and deletion policies, and adherence to data handling standards required by frontier lab partners