Model Calibration Specialist

Doordash•Tempe, AZ

51d•Hybrid

About The Position

As a Model Calibration Specialist, you'll be a key player in scaling high-quality QA outcomes across DoorDash Support. You will serve as a bridge between human QA reviewers, machine learning systems, and cross-functional stakeholders. Your mission: ensure model-generated QA outcomes are accurate, fair, and reflect real-world expectations. You'll review conflicts between human and model decisions, investigate edge cases, and surface opportunities for improvement. This work will directly impact how we calibrate our ML model, refine training data, and optimize support agent workflows. You'll report to the Supervisor of the Model Calibration Team (MCT) within the CXI Quality Assurance organization. This is a hybrid role based in Tempe, AZ, requiring in-office presence based on business needs, currently one day per week. Candidates must live within 50 miles of the Tempe, Arizona, DoorDash Corporate Office.

Requirements

Experience: 2+ years in Quality Auditing, or a related role.
Communication: You can clearly convey insights and escalate edge cases across teams, from frontline agents to senior stakeholders.
Analytical Thinking: You enjoy working from first principles, testing hypotheses, and validating ideas with data.
Problem Solving: You thrive on ambiguity, translating blockers into actionable steps and scalable improvements.
Independence: You can manage your own queue and priorities in a fast-paced, team-oriented environment.
Tools: Proficient in Microsoft and/or Google Suite
Mindset: You bring an ownership mentality, strive for continuous improvement, and embody the 1% better philosophy.

Nice To Haves

a plus if you have experience with machine learning processes.

Responsibilities

Develop deep expertise in DoorDash's support systems and workflows to support better model calibration and review alignment.
Review and resolve conflicts between ML model outputs and human QA decisions, especially in ambiguous or high-risk cases.
Analyze patterns in false positives, false negatives, and overrides to identify areas for calibration or reviewer training.
Act as a subject matter expert for edge case resolution and decision documentation.
Recommend improvements to agent workflows and support tools based on calibration findings.
Contribute directly to improving the training data and feedback loop used by our automated QA system - ensuring our model reflects human standards of quality and fairness.