Want to know if this job is worth applying to?
San Francisco, California, United States
Hybrid
At Bedrock, we're moving AI out of the lab and into the real world. Our team includes veterans who helped launch Waymo, scaled Segment to a $3.2B acquisition, and grew Uber Freight to $5B in revenue. Today, we're deploying autonomous systems on heavy construction equipment across the country, improving safety on job sites and accelerating schedules on critical infrastructure projects.
We're not here debating the future of AI. We're deploying it in the real world. In just two years, we've raised $350M and achieved the first fully autonomous excavator deployments in construction.
This is where algorithms meet steel-toed boots. You'll work alongside construction veterans and world-class engineers to solve physical-world problems that simulations can't touch. If you're ready to do meaningful work on hard problems, we'd love to have you join us.
Bedrock is building autonomous construction robots that operate in some of the most unstructured, high-stakes environments on earth. To do that, we need machines that deeply understand the built world - and that starts with data.
The Data Understanding Program is the backbone of how we annotate, search, and explore the robot data that powers our autonomy stack. As an Agentic Data Understanding Engineer, you will design and build the AI-driven systems that automatically generate high-quality annotations at scale: the auto-labelers, orchestration pipelines, quality harnesses, and agentic workflows that allow Bedrock to move from raw sensor data to structured, semantically rich annotations - quickly, cheaply, and reliably.
This role sits at the intersection of applied ML, data infrastructure, and robotics, and will directly shape how fast Bedrock can expand its operational domain.
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.
Develop and maintain annotation harnesses that automatically assess quality and cost against golden/validation sets to enable continuous hill-climbing on label quality.
Productionize annotation workflows: backfilling historical data, triggering runs on new data programmatically, and tracking coverage across the fleet.
Build and maintain the Annotation Orchestrator that executes annotation job sequences, tracks progress, and manages integrations with external vendors.
Implement scheduling and prioritization logic for annotation jobs across multiple parallel workflows and data modalities (2D boxes, 3D cuboids, events, metrics, embeddings).
Own the annotation job configuration library - the registry of available annotators - and ensure new annotator technologies are onboarded in a scalable, secure way.
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.
Develop tooling for annotation versioning and downstream propagation so that spec updates automatically trigger re-annotation of affected data.
Partner with the Data Infrastructure team to store annotations in queryable, version-controlled databases that support cross-domain semantic search and exploration.
Define and instrument the north star metrics for annotation quality and cost.
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.
Identify and evaluate new annotation technologies.
8+ years of experience in ML engineering, data engineering, or applied AI — with meaningful time spent building production annotation or labeling systems.
Strong experimentation coding skills (eg Python); experience building complex pipelines that include various ML model inference at scale.
Hands-on experience with one or more annotation modalities: 2D bounding boxes, 3D cuboids, semantic segmentation, event/context labels.
Clear thinking about quality/cost tradeoffs in data pipelines; comfort defining and instrumenting metrics.
Ability to work cross-functionally with autonomy engineers, product managers, and human labeling teams.
Bonus: Hands on model fine tuning experience.
Experience building agentic or LLM-orchestrated pipelines (e.g., using VLMs for zero-shot or few-shot annotation).
Familiarity with robotics data formats and sensor modalities (LiDAR, cameras, IMU).
Experience with annotation platforms.
Background in robotics, autonomous vehicles, or construction technology.
Experience designing ontologies or taxonomies for structured data labeling.
Bedrock is automating construction - one of the largest, most dangerous, and least digitized industries in the world. Our robots work on real job sites today, and we’re scaling fast. The data you help annotate and understand will directly train the perception and autonomy systems that keep workers safe and accelerate project timelines.
You’ll work with a tight-knit team of engineers who care deeply about the quality of what we build, have access to rich real-world data from our fleet, and own meaningful surface area from day one.
Our roles are often flexible. If you don't fit all the criteria, or are in another location (especially one where we have an office like SF or NY), please apply anyway! We'd love to consider you.
Bedrock Robotics is an Equal Opportunity Employer
We’re committed to building a diverse and inclusive workplace. We consider all qualified applicants for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, ancestry, age, disability, veteran status, genetic information, or any other protected characteristic.
Reasonable Accommodations
We want our hiring process to be accessible to everyone. If you need an accommodation to participate in the application or interview process, please let your recruiter know so we can support you.
