The Ampin App
deep insights into the Atlassian community

As a Staff Data Scientist at Zendesk, you will join the AI Security team to envision and develop machine learning and AI approaches that help detect anomalous activity and AI security risks, collaborating with Security Engineers, Software Engineers, ML researchers, and product teams to share actionable insights and models that keep Zendesk's customers safe and prevent product abuse. Your key responsibilities include running and developing the AI Security evaluation platform, analyzing large volumes of behavioral and user interaction data to identify abuse patterns and automate methodologies, developing detection models from offline data to guide production, improving spam detection accuracy while minimizing false positives, and collaborating with cross-functional teams to implement prevention strategies. Required qualifications include an advanced degree or equivalent in a quantitative field, 5+ years of ML/AI experience in fraud, anomaly detection, or security, strong programming skills in SQL and Python with data processing framework proficiency, experience with AI-assisted programming tools, and expertise in a range of statistical and ML techniques, particularly anomaly detection on unbalanced data, plus experience with large datasets and distributed computing environments like Spark and Snowflake and strong communication skills. Preferred qualifications include a proven track record implementing anomaly detection or AI Security evaluation systems at scale, experience with AI red-teaming tools, cybersecurity knowledge of common attack vectors, and familiarity with ML model monitoring and production guardrails and customer impact considerations. Zendesk emphasizes a hybrid work model, a commitment to fairness and inclusion, and transparency about using AI or automated decision systems to screen applicants; the company is an AA/EEO/Veterans/Disabled employer and offers accommodations for applicants with disabilities, with instructions to request accommodations via email.