Data Scientist, Decision Quality & Impact

Sperry Rail, Inc.; Shelton (CT); On-site; USD 100,000 - 160,000; Full time

Job summary

Sperry Rail, Inc. in Shelton, CT is seeking a Data Scientist focused on decision quality. You will model the decisions of analysts reviewing neural-network–identified defects from ultrasonic and electromagnetic data, aiming to improve labeling accuracy and training thresholds.

This role blends data science with domain knowledge of rail inspection and risk analysis. You will join a small US team, design retrospective studies, quantify inter-analyst variation, and produce actionable insights for

Qualifications

• Strong Python, NumPy, Pandas, Scikit-learn, and SQL proficiency.
• Applied statistics, classification metrics, sampling, and bias assessment.
• Experience with imbalanced and imperfectly labeled data.
• Ability to explain methods to non-technical stakeholders.
• Familiarity with Git, CI/CD, and Agile practices.
• Bachelor's degree in statistics, CS, engineering, or related field.
• Rail testing, NDT, or sensor-based inspection exposure.

Responsibilities

• Build retrospective scoring of analyst dispositions and assess decision quality.
• Quantify agreement across analysts, shifts, and conditions.
• Identify factors behind missed or misinterpreted defects and create a usable taxonomy.
• Develop feedback loops—training, tooling, and threshold recommendations.
• Define measures of decision quality and monitor over time.
• Present findings to leadership and engineers; write clean, tested code.
• Participate in code reviews, sprint planning, and documentation.
• Work with detection and platform teams to ensure signals are captured.

Skills

Python

SQL

Statistics

Imbalanced ML

Ground truth

Communication

Git & CI/CD

Agile

Human-in-the-loop

Rail sensors

Rail industry

Education

Bachelor's degree in quantitative field

Tools

Git

CI/CD

Job description

Sperry Rail, Inc. in Shelton, CT is seeking a Data Scientist focused on decision quality. You will model the decisions of analysts reviewing neural-network–identified defects from ultrasonic and electromagnetic data, aiming to improve labeling accuracy and training thresholds.

This role blends data science with domain knowledge of rail inspection and risk analysis. You will join a small US team, design retrospective studies, quantify inter-analyst variation, and produce actionable insights for