Work Experience

June 2025 – Present

Product Overview: Contributed to the Qianfan AppBuilder—a one-stop, AI-native application development workbench built on large language models.

• Agent Capability Enhancement: Drove iterative improvements in our AI agents' performance by designing and executing evaluation frameworks across data quality, model accuracy, and interaction flows; delivered industry-specific benchmark templates for rapid customization.
• User Experience Optimization: Spearheaded initiatives to lower the barrier to entry and boost customization options, significantly enhancing the product experience for users building AI-powered applications.

April 2025 – June 2025

Product Overview: Contributed to Ark (Fangzhou), Volcano Engine's full-lifecycle large-model service platform—for training, inference, evaluation, and fine-tuning.

• User-Needs Analysis: Analyzed support tickets, survey feedback, and usage metrics to identify high-frequency pain points and defined a roadmap for documentation improvements.
• Technical Documentation: Partnered with product and engineering teams to design and author comprehensive API references, quick-start guides, and best-practice tutorials for large-model integration.
• AI-Tool Co-Development: Collaborated on the development of a documentation Q&A chatbot, an automated error-checking utility, and a PRD-content understanding plugin—streamlining our documentation workflow and boosting developer productivity.

August 2024 – February 2025

Product Overview: Vertical-Industry Environmental LLM & AI-Powered Bidding Agent

• Project 1: Vertical-Industry Environmental LLM
• Challenge: Enterprises faced 15-day response times and low accuracy in legal citations when querying government mailboxes with general-purpose LLMs.
• Solution: Constructed a 'Bad Case' dataset and a multi-dimensional evaluation framework; ingested and structured unstructured Q&A data to build a legal knowledge base.
• Optimization: Systematically refined prompt engineering and RAG retrieval strategies; led model selection and fine-tuning efforts.
• Results: Achieved a +0.30 lift in F1 score and +0.25 in GPT-based evaluation; slashed response latency from 15 days to 1 minute, enabling rapid deployment of a vertical-scenario LLM product.
• Project 2: AI-Powered Bidding Agent
• Goal: Address low efficiency and extraction accuracy in the bidding process.
• Approach: Designed a multi-agent system for bid-document understanding, generation, and automated scoring; systematically delineated LLM capabilities in structured parsing vs. free-form generation.
• Achievements: Extracted key points from 100 pages in under 5 minutes with 92% accuracy and 95% coverage—dramatically improving corporate bidding efficiency and proposal quality.

April 2024 - August 2024

• Analyzed financial documents and earning calls using NLP to derive data-driven conclusions.

December 2023 - April 2024

• Conducted in-depth industry research on upstream and downstream sectors from annual reports, financial news to support corporate strategic planning and investment decisions on industrial robotics.

October 2022 - January 2023

• Participated in carbon emissions inventory project for Hechuan District, Chongqing City. Calculated emissions from industrial, agricultural, and forestry sectors using the emission factor method.