LLM Evaluation Engineer

Job Description

• Build and maintain automated LLM evaluation pipelines covering multiple dimensions, including general capabilities, Agent capabilities, and persona/role-playing capabilities. Support one-click evaluation, historical result comparison, and regression testing.
• Conduct general capability evaluations using established benchmarks such as MMLU, C-Eval, HumanEval, GSM8K, MATH, and IFEval, including benchmark deployment, execution, and results analysis.
• Conduct Agent capability evaluations, including setting up evaluation environments and tracking performance metrics for benchmarks such as BFCL, τ-bench, and GAIA.
• Design and execute persona and role-playing evaluation frameworks, covering metrics such as identity recognition, role consistency, multi-turn stability, and style consistency.
• Record and analyse evaluation results across model training runs, perform comparisons and anomaly detection, and prepare evaluation reports for training checkpoints.
• Conduct regular intermediate evaluations during the pre-training stage to track the evolution and improvement of model capabilities.

Job Requirements

• Bachelor's degree or above in Computer Science, Artificial Intelligence, or a related discipline.
• Familiar with mainstream LLM evaluation benchmarks and frameworks, such as lm-eval-harness, OpenCompass, and EvalPlus.
• Proficient in Python, with the ability to independently build evaluation pipelines covering model inference/deployment, batch evaluation, and results analysis.
• Familiar with LLM inference frameworks such as vLLM and SGLang, with the ability to deploy models for batch inference and evaluation.
• Experience in evaluation data analysis and visualisation.
• Detail-oriented and rigorous, with a strong focus on ensuring the reproducibility and reliability of evaluation results.

Good to Have

• Experience with Agent evaluation, particularly BFCL, τ-bench, GAIA, or SWE-bench.
• Experience with persona or role-playing evaluation, such as CharacterBench or RMTBench.
• Experience in evaluation automation and CI/CD integration.
• Understanding of LLM training workflows, with the ability to interpret the relationship between training checkpoints and evaluation results.

Skills

Recognition, Reporting Training, Regression Testing, Mathematics, Pipeline Development, scientific discipline, Artificial Intelligence, Batch Processing, Data Evaluation, Inference Engine, Capability Development, Reproducibility, Role Playing, Capability Analysis, Benchmarking, Performance Evaluations