Job Summary:
Bedrock Robotics is a company focused on bringing advanced autonomy to the construction industry. They are seeking an Agentic Data Understanding Engineer to design and build AI-driven systems for high-quality data annotation, which is crucial for their autonomous construction robots. This role involves working at the intersection of applied machine learning, data infrastructure, and robotics to enhance the operational capabilities of their technology.
Responsibilities:
โข Design and build a hybrid cascading auto-labeling pipeline that intelligently selects annotation techniques - onboard sensor-derived labels, specialized AI models, and VLMs with prompt/few-shot configs - based on quality, cost, and latency requirements.
โข Develop and maintain annotation harnesses that automatically assess quality and cost against golden/validation sets to enable continuous hill-climbing on label quality.
โข Productionize annotation workflows: backfilling historical data, triggering runs on new data programmatically, and tracking coverage across the fleet.
โข Build and maintain the Annotation Orchestrator that executes annotation job sequences, tracks progress, and manages integrations with external vendors.
โข Implement scheduling and prioritization logic for annotation jobs across multiple parallel workflows and data modalities (2D boxes, 3D cuboids, events, metrics, embeddings).
โข Own the annotation job configuration library - the registry of available annotators - and ensure new annotator technologies are onboarded in a scalable, secure way.
โข Build agentic annotation workflows that can reason about annotation quality gaps, self-diagnose failures, trigger human-in-the-loop review when confidence is low, and iterate toward coverage goals autonomously.
โข Develop tooling for annotation versioning and downstream propagation so that spec updates automatically trigger re-annotation of affected data.
โข Partner with the Data Infrastructure team to store annotations in queryable, version-controlled databases that support cross-domain semantic search and exploration.
โข Define and instrument the north star metrics for annotation quality and cost.
โข Build reporting and dashboards to surface quality/cost trends across annotation workflows and inform decisions on when to invest in better auto-labeling vs. human annotation.
โข Identify and evaluate new annotation technologies.
Qualifications:
Required:
โข 8+ years of experience in ML engineering, data engineering, or applied AI โ with meaningful time spent building production annotation or labeling systems.
โข Strong experimentation coding skills (eg Python); experience building complex pipelines that include various ML model inference at scale.
โข Hands-on experience with one or more annotation modalities: 2D bounding boxes, 3D cuboids, semantic segmentation, event/context labels.
โข Clear thinking about quality/cost tradeoffs in data pipelines; comfort defining and instrumenting metrics.
โข Ability to work cross-functionally with autonomy engineers, product managers, and human labeling teams.
Preferred:
โข Hands on model fine tuning experience.
โข Experience building agentic or LLM-orchestrated pipelines (e.g., using VLMs for zero-shot or few-shot annotation).
โข Familiarity with robotics data formats and sensor modalities (LiDAR, cameras, IMU).
โข Experience with annotation platforms.
โข Background in robotics, autonomous vehicles, or construction technology.
โข Experience designing ontologies or taxonomies for structured data labeling.
Company:
Bedrock Robotics enables the construction sector to grow at the rate that modern civilization requires. Founded in 2024, the company is headquartered in San Francisco, USA, with a team of 51-200 employees. The company is currently Growth Stage.