The opportunity
Binance is a leading global blockchain ecosystem behind the world’s largest cryptocurrency exchange by trading volume and registered users. We are trusted by 300+ million people in 100+ countries for our industry-leading security, user fund transparency, trading engine speed,…
What you'll do
Design end-to-end LLM evaluation plans for business scenarios such as: dialogue and financial trading. Build evaluation metric systems and rubrics, transforming subjective model performance judgments into quantifiable, reproducible, and explainable evaluation conclusions.
Lead the design and construction of evaluation datasets. Define evaluation: dimensions and scenario coverage, establish high-quality data annotation guidelines and quality control processes, and build benchmarks that authentically reflect business needs and have discriminative power.
Analyze model capability boundaries and failure modes based on evaluation: results. Produce actionable improvement recommendations and collaborate with algorithm and product teams to drive model iteration, making evaluation a critical component of the R&D loop.
Drive the automation and scaling of evaluation workflows. Build sustainable: evaluation platforms and toolchains to support high-frequency, stable evaluation needs during rapid model iteration.
Collaborate with algorithm, product, and data teams to translate business and: model objectives into clear evaluation standards, and turn evaluation findings into concrete R&D directions and drive their implementation.
What they're looking for
- Master's degree or above in Computer Science, Artificial Intelligence,: Mathematics, Statistics, or related fields, with a solid algorithmic foundation and understanding of LLM principles, training, and fine-tuning processes.
- Hands-on LLM evaluation experience at a large tech company, with: participation in commercial deployment evaluation (not purely academic or offline benchmarking). Familiar with the full pipeline from evaluation data preparation and rubrics design to evaluation-driven R&D.
- Familiar with mainstream evaluation methods (human evaluation, model-based: automatic evaluation / LLM-as-a-judge, metric computation) and their applicable boundaries. Able to define appropriate evaluation dimensions for different business scenarios and write clear, actionable, and discriminative rubrics.
- Systematic control over evaluation data representativeness, annotation: consistency, and result reliability, ensuring scientific and trustworthy evaluation conclusions.
- Proficient in Python, with experience in evaluation workflow automation,: benchmark construction, or evaluation platform development. Able to independently handle data processing, evaluation script writing, and result analysis.
- Strong business understanding and communication skills, able to translate: evaluation findings into clear improvement directions and effectively drive cross-team collaboration.
