数据科学家（模型评估方向）-TikTok
Core
Build end-to-end evaluation systems for content safety models (ASR, OCR, LID, traditional ML+LLM) to support model capability, launch decisions, and continuous iteration.
Role type
Senior IC data scientist (model evaluation)
Builds
Content safety models and evaluation frameworks
Domain
Content safety / AI / GenAI
Deliverable
production ML models
Required skills
LLM evaluation, multi-modal model assessment, evaluation set construction, data cleaning, risk control modeling, causal analysis, experimental design, SQL, Python
Preferred skills
Agentic AI data solutions, prompt evaluation, automated evaluation pipelines, model drift detection, RCA analysis
Technologies
SQL, Python, LLM, ASR, OCR, LID
Responsibilities
Design training and evaluation sets for multi-language/multi-modal scenarios; establish model launch standards and acceptance mechanisms; build post-launch monitoring and alerting systems; drive model optimization via root cause analysis; standardize evaluation methodologies.
Seniority
Senior, hands-on IC
