**About the Role** We are looking for a driven and detail-oriented **L5B Program Manager** to join our **Frontier Labs AI team**, focused on building high-quality, multi-modal data pipelines to support advanced model development and foundational research. In this role, you will lead the **end-to-end execution** of AI data labeling workflows across **text, image, audio, video, and instruction-tuned datasets**, partnering closely with researchers, data scientists, product managers, and annotation vendors. You will play a critical role in **scaling and operationalising labeling operations**, ensuring that the data used to train and evaluate cutting-edge models is accurate, diverse, and aligned with evolving research needs. This is a hands-on role for someone who thrives in **high-ambiguity, high-velocity environments** and can bring structure and discipline to rapidly evolving labeling workflows **\-\-\-\- What You Will Do ----** ### **Program Execution & Delivery** 1. Manage AI data labeling programs from **scoping to delivery**, ensuring high-quality annotations at scale. 2. Translate **Frontier Labs research needs** into concrete annotation specs, rubrics, and task designs. 3. Own timelines, throughput plans, and quality controls for critical datasets used in LLM training and evaluation. ### **Stakeholder Management** 1. Partner with researchers, data scientists, product, and ops to ensure labeling goals are aligned with model objectives. 2. Work cross-functionally to drive task clarity, resolve ambiguity, and incorporate feedback into successive batches. 3. Act as the **single-threaded owner** for specific labeling programs, managing internal and external partners. ### **Operational Infrastructure** 1. Develop and refine **batching strategies**, **smart sampling plans**, and **audit workflows**. 2. Drive **QA processes**, including golden set calibration, rubric refinement, and disagreement adjudication. 3. Ensure traceability from **raw inputs to final labeled outputs**, and track quality regressions over time. ### **Process Design & Automation** 1. Identify opportunities to apply **model-in-the-loop labeling**, **active learning**, or **self-checking pipelines**. 2. Collaborate with tool owners and engineers to integrate annotation workflows with internal tooling systems. 3. Own feedback loops that enable raters to improve over time and reduce error variance **\-\-\-\- What You Will Need  ----** **Bachelor’s degree** in Engineering, Data Science, Linguistics, or related technical/analytical field. **5+ years** of program or project management experience in AI/ML, data ops, or labeling infrastructure. Demonstrated ability to manage **end-to-end data pipelines** in AI/ML or research environments. Strong working knowledge of **Robotics, Physical AI Data labeling tasks**, such as: 01. Object detection and recognition 02. Semantic & Instance Segmentation 03. Depth & Pose Estimation 04. Grasp Detection 05. Action Segmentation 06. Trajectory Labeling 07. Prompt-response evaluation 08. Instruction tuning 09. Dialogue evaluation 10. Vision-language QA 11. Video slot tagging 12. Image Tagging 13. Documentation Extraction 14. Data collection annotation 15. HRI Experience collaborating with research or model teams to scope data collection requirements. Excellent written and verbal communication skills \-\-\-\- Preferred Qualifications ---- 1. Experience in **frontier AI research environments**, such as foundation model labs or GenAI startups. 2. Familiarity with tools like **Label Studio, Scale AI, SuperAnnotate, Snorkel Flow, or in-house annotation platforms**. 3. Understanding of LLM training and evaluation lifecycles. 4. Experience working with **human-in-the-loop systems** or model-assisted labeling pipelines. 5. Familiarity with **multilingual or multi-cultural annotation programs** Uber's mission is to reimagine the way the world moves for the better. Here, bold ideas create real-world impact, challenges drive growth, and speed fuelds progress. What moves us, moves the world - let’s move it forward, together. Offices continue to be central to collaboration and Uber's cultural identity. Unless formally approved to work fully remotely, Uber expects employees to spend at least half of their work time in their assigned office. For certain roles, such as those based at green-light hubs, employees are expected to be in-office for 100% of their time. Please speak with your recruiter to better understand in-office expectations for this role. \*Accommodations may be available based on religious and/or medical conditions, or as required by applicable law. To request an accommodation, please reach out to [accommodations@uber.com](mailto:accommodations@uber.com).