Data Pipeline Lead, Player One

About the Operation This role sits inside VNG's egocentric physical AI data collection operation, a joint venture born from VNG's strategic partnership with a leading global hardware partner. We collect large-scale, real-world data from factory floors, hotels, logistics centres, and beyond, and deliver it as training-ready datasets to frontier AI customers including OpenAI Lab. Our stack: partner head-mounted cameras and tactile gloves for capture; our US technical partner for data enrichment and pipeline methodology; VNG's operational reach in Vietnam and Southeast Asia for scale. Target: 400,000+ qualified hours by April 2027 from a standing start. The pipeline that makes that possible does not yet exist. You will build it.

About the Role You will own the end-to-end data pipeline from device capture to customer delivery. This is a founding technical hire. You design the architecture, stand it up, hire the team that runs it, and keep it healthy as volume grows from hundreds to hundreds of thousands of hours per month. You sit at the intersection of hardware operations, software engineering, and AI data quality. You work directly with the hardware partner's Shenzhen team on platform integration, with our US technical partner on enrichment methodology, and with the Vietnam operations team on scenario planning and factory deployment.

What You Will Do 1. Pipeline Architecture & Build Design the full pipeline from device capture to customer delivery: sync, ingestion, processing, QC, labeling, format conversion, export. Define storage architecture to handle 20-25 GB/hour per device and scale to petabytes across millions of collected hours. Build or integrate automated processing stages: AI segmentation, task labeling, privacy blurring (face/plate anonymisation), quality scoring. Establish data delivery formats: LeRobot, VLA, video-generative, and custom buyer formats.

1. Quality Control Design and implement automated QC: completeness checks, blur/motion-artefact detection, label consistency, qualification scoring. Build manual review workflows for borderline episodes and new scenario types. Define and maintain 'qualified hour' standards -- the threshold that determines what gets delivered and billed. Track qualification rates across factories, scenarios, and device configs; feed results back into operations.
2. Device & Sync Integration Establish the device-to-cloud data path (Bluetooth to phone to cloud), handling large transfers under field conditions. Manage data card protection, batch ingestion pipelines, and device management integrations. Collaborate with the mobile engineering team on sync architecture and the collector-side app.
3. Partner Integration Interface with the hardware partner's collection platform: device management, task creation, automatic QC hooks, Vietnam-scenario customisation. Collaborate with our US technical partner on enrichment: trajectory segmentation, reward signal generation, RL-ready dataset conversion. Coordinate with the hardware partner's Shenzhen team on hardware specs and processing benchmarks.
4. Team & Operations Leadership Build and lead the data pipeline team (engineers + QC operators); prepare headcount plan against 2027 budget. Define and document all pipeline SOPs; keep them current as collection scenarios expand. Report pipeline health, throughput, and quality metrics weekly to the Head of PT Lab 5+ years in data engineering, ML data infrastructure, or a closely related field. Strong in Python, SQL, cloud object storage (S3-compatible), and large-scale data processing pipelines. Experience with video data pipelines; familiarity with CV annotation tooling (CVAT, Label Studio, or similar). Comfortable with edge-to-cloud data flows: IoT/mobile sync, large file transfer at scale. Demonstrable experience defining and enforcing data quality standards in a production environment. Hands-on team leader -- willing to code and to manage. Professional English; Mandarin or Vietnamese is a strong plus. Strongly Preferred Experience with robotics or embodied AI data formats: LeRobot, HDF5, rosbag, ACT, OpenVLA, RLDS. Familiarity with egocentric or teleoperation datasets (Ego4D, EPIC-Kitchens, Open X-Embodiment). Experience with privacy/anonymisation pipelines (GDPR-class standards). Experience building B2B data delivery workflows or data marketplaces. Background in autonomous vehicles, physical AI, or high-frequency multi-modal sensor fusion *** Why This Role Founding pipeline hire -- design from scratch, not inherit legacy architecture. JV backed by VNG, Vietnam's largest internet company, and a leading global hardware partner. Direct pipeline to frontier AI customers including OpenAI Lab. 0 to 400K+ qualified hours in six months -- build at real scale, fast. Based in HCMC with regular collaboration in Shenzhen and San Francisco