Work Experience
工作经历
Co-founded an AI-driven compliance automation startup for Testing, Inspection and Certification (TIC) workflows, developing systems for regulatory reasoning, report generation, and customer management. Our system combines hybrid document retrieval, regulatory knowledge graphs, and structured report generation with traceable evidence, while using a standard-agnostic architecture to support rapid adaptation across different national standards and product categories. We also developed automated customer-management workflows and supported early commercial deployment with international TIC companies including Intertek and DEKRA.
- Co-founded an AI-driven compliance automation startup for Testing, Inspection and Certification (TIC) workflows, developing systems for regulatory reasoning, report generation, and customer management.
- Built machine-learning-based compliance reasoning systems combining hybrid document retrieval, regulatory knowledge graphs, and structured report generation with traceable evidence.
- Designed standard-agnostic architectures that separate regulatory clauses from workflow logic, enabling rapid adaptation across different national standards and product categories.
- Developed an automated annual-fee management system to summarize customer service records, calculate annual fees, generate client quotations, and streamline email-based customer communication.
- Supported early commercial deployment with international TIC companies including Intertek and DEKRA, validating the system through real industrial compliance-reporting and customer-management workflows.
Developed HDCoin, a proof-of-useful-work blockchain framework that transforms conventional mining into useful training and verification of trustworthy biomedical Hyperdimensional Computing (HDC) models. The project combines distributed HDC model training, verification, design-space exploration, and adaptive mining mechanisms across multiple biomedical learning tasks.
- Developed HDCoin, a proof-of-useful-work blockchain framework that transforms conventional mining into useful training of trustworthy biomedical Hyperdimensional Computing (HDC) models.
- Designed HDC-based mining and verification pipelines where distributed miners train, validate, reproduce, and verify biomedical learning models using nonce-generated item memories and HDC hyperparameter configurations.
- Conducted design-space exploration across HDC dimensionality, retraining iterations, and retraining rates on four biomedical datasets, including UCIHAR, BreastMNIST, PneumoniaMNIST, and NoduleMNIST3D.
- Evaluated adaptive mining difficulty and fairness mechanisms, showing how HDC configurations affect computational effort, model performance, and miners’ winning probability.
Supported autonomous driving R&D through cross-functional project coordination, rule-based obstacle avoidance and emergency decision-making algorithm development, and 3,000+ miles of real-world highway testing. Translated road-test observations and customer needs into actionable engineering and product requirements.
- Supported end-to-end project management for autonomous driving R&D projects, coordinating algorithm, software, and testing teams while tracking milestones, development schedules, and cross-functional resources. Translated road-test issues into actionable algorithm requirements and engineering tasks to improve development, testing, and delivery efficiency.
- Contributed to the development of rule-based obstacle avoidance and emergency decision-making algorithms for high-risk scenarios involving water barriers, traffic cones, road obstacles, and sudden vehicle cut-ins. Designed decision logic linking obstacle detection results to vehicle behavior and maneuver selection.
- Participated in real-world autonomous driving tests on elevated highways in Shanghai, completing 3,000+ miles of on-road testing in real traffic conditions. Analyzed abnormal scenarios, tuned rule thresholds, and validated algorithm performance by investigating false triggers, missed responses, and unstable strategy transitions, contributing to improved emergency avoidance rules and broader scenario coverage.
- Served as a liaison among product teams, engineering teams, and external clients, translating business and customer needs into actionable technical and product requirements while supporting solution integration and delivery.