Sr. Staff Autonomy Product Manager
Job Description
Rivian is building autonomy at fleet scale, and this role owns the product vision, strategy, and roadmap behind autonomy data collection and scenario search/curation tooling. You will lead an end-to-end loop where fleet data is triggered, curated, indexed for retrieval, and then used to improve model training and simulation outcomes.
This is a hands-on product role that connects engineering needs across cloud search, onboard logging triggers, ML engineering, and scenario review workflows, while driving loop performance with measurable quality and operational metrics.
What you will do
- Own the vision, strategy, and multi-quarter roadmap for data collection and curation tooling, tying each bet to a model-quality outcome.
- Write user stories, requirements, and acceptance criteria that cloud, onboard logging, and ML engineering teams can implement.
- Drive the loop end to end: define “we need more X,” find existing examples, trigger written and validated capture on the right vehicles, then ensure data is uploaded and indexed, curated, and ultimately used in training or simulation with impact measured.
- Lead embedding-based search including text-to-scene and example-to-scene retrieval, similarity search over fleet video and sensor data, and clustering to surface new or rare scenarios and remove near-duplicates.
- Bring cloud search and onboard triggers into one shared embedding space so a scene found in the cloud can become a fleet trigger and the triggered data flows straight back into search.
- Define and track loop metrics such as retrieval precision and recall, trigger yield, upload cost per useful sequence, time from request to curated dataset, reviewer throughput, and dataset impact on model regressions.
- Improve curation speed and trust using automated QA with vision-language models, efficient review workflows, and enrichment including tags, metadata, and scene descriptions.
- Stay hands-on by querying data, reviewing scenes, and running search-quality and yield experiments.
- Translate needs across perception, planning, end-to-end model, and simulation teams into a roadmap and user stories for two connected product areas.
Two connected product areas
- Fleet data collection (dynamic triggers): engineers start from a sequence, a dataset, or a scene description to put customer and test-fleet vehicles “on the lookout” for similar situations. These triggers are deployed over the air without waiting for a software release.
- Scenario search and curation platform: the system turns a plain-language request or an example clip into a checked, human-reviewed dataset, extending beyond keyword and metadata search to embedding-based search across the full fleet data corpus.
What you bring
- A record of shipping data infrastructure, search or retrieval products, or data-centric AI tooling used by engineers and researchers.
- Working knowledge of embeddings, vector search, and multimodal models, including how retrieval quality is measured.
- A solid grasp of how training-data quality, diversity, and long-tail coverage drive model performance.
- Ability to produce specifications and user stories aligned with internal customer needs and outcomes that are outsize relative to effort.
- The ability to influence many teams and senior leadership without direct authority.
Technologies
- embeddings, vector search, multimodal models, vision-language models
Nice to have
- Data experience in autonomous driving, robotics, or ADAS including fleet logging, triggers, scenario mining, auto-labeling, or simulation.
- Experience with open source LLMs and VLMs and how they can enhance a SoTA data loop.
- Privacy-by-design and consent handling for customer data.
- A degree in CS, EE, or robotics, or equivalent experience.
Location: Palo Alto, CA (onsite)