Experience

Just Horizons Alliance (JHA)

Benchmark Research AssistantJune 2025 – January 2026 · Boston, MA
AI GovernanceAI BenchmarkAPI

Worked on the development of an ethics-oriented LLM evaluation framework, the AI Ethics Index (AIEI), addressing limitations in existing benchmarks regarding ethical behavior and governance compliance, covering safety, model design, environment impact, fairness, transparency, etc. using a tree architecture.

Built end-to-end evaluation pipelines supporting both local (LLaMA) and API-based (Gemini) models, automating prompt testing and data analysis. See the following flow chart for detailed procedure. Additional temperature-tiered testing are executed to enhance the richness and adding robustness.

For additional analysis and existing reports regarding financial coercion and doxxing patterns, click here.

Evaluation pipeline

Institute of Automation, Chinese Academy of Sciences

NLP Research AssistantSept 2023 – March 2024 · Beijing, China
LLM FinetuningLangchainRAGPrompt Engineering

Worked in a Multi-modal Biological Experiment Agent project, with a focus on natural language processing and system integration. The system was designed to support both automated experiment execution and knowledge-assisted searching in laboratory workflows

Developed a Knowledge Base QA system with 20,000+ biological protocols using BeautifulSoup and Markdown for efficient retrieval. The answering system enables querying of protocol procedures, step-specific guidance, and contextual recommendations on experimental focus and execution details. It was integrated into a complete web-based user interface.

Engineered a LangChain-based Agent translating CV outputs into robotic commands. To improve robustness and controllability, an action–task mapping dictionary is designed to constrain model outputs and reduce hallucinations. Through model fine-tuning, prompt engineering, few-shot learning, chain-of-thought prompting , and the constrained mapping layer, the system achieved 98% execution stability and a 57% reduction in end-to-end latency.

System UI

Snowball

Big Data Development InternFeb 2023 – May 2023 · Beijing, China

Identified and corrected data integrity issues across 10 warehouse tables and 68 sensitive fields using SQL, while optimizing warehouse logic by removing redundant dependencies.

Reviewed 211 ODS tables and 681 downstream tables, identifying two unnecessary dependencies and more than one hundred instances of redundant logic.

Bianlifeng Technology

HRIS & Data Analytics InternJune 2021 – Sept 2021 · Beijing, China

Developed RPA workflows improving operational efficiency by 95% and processed large-scale employee data using SQL window functions for multi-department reporting. In one project, the system automatically extracted and exported information for more than 8,400 retired employees for Uber billing coverage.

Conducted data analysis using SQL to extract information from the data warehouse and applied window functions to process and categorize employee work data, supporting recurring monthly reporting needs for more than 10 departments over multiple reporting cycles.