Xinai Lu

Research Assistant | Feb 2026 – Present

Focusing on domain-specific LLM system and Multi-Agent system design (NIH Stage Model).

Research Assistant | Jun 2024 – May 2025

Focused on CV & Multimodal in OOD detection scenarios.

Du, Pan; Zhao, Wangbo; Lu, Xinai; Liu, Nian; Li, Zhikai; Gong, Chaoyu; Zhao, Suyun; Chen, Hong; Li, Cuiping; Wang, Kai; You, Yang. “Unsupervised Learning for Class Distribution Mismatch.” Proceedings of the 42nd International Conference on Machine Learning (ICML 2025), PMLR 267:14682-14718, July 2025.

Data Science Intern | May 2025 – Aug 2025

Applied causal inference methods (DML, DR-Learner, PSM, GAM) to estimate treatment effects of pricing strategies. Built driver–city panel data and achieved ~5.2% RMSE reduction with DR-Learner + XGBoost. Evaluated time-series policies using Google Causal Impact when A/B testing was infeasible.

Statistician Intern | Jan 2024 – Feb 2024

Conducted statistical modeling and clinical trial analysis. Built an internal analysis platform using R Shiny to improve experimental efficiency and workflow automation.

Designed a reinforcement learning framework using execution-based rewards (precision, recall, F1, IoU). Built a dual-loop optimization system combining LLM-based reward generation (Eureka-style) and PPO policy training.

Built a volatility prediction pipeline combining FinBERT embeddings, RAG-enhanced financial text, and LLM-extracted semantic signals. Designed cross-firm and temporal attention mechanisms, achieving ~0.70% MSE improvement.

Modeled causal relationships using Double Machine Learning (DML) and interpretable ML methods. Analyzed the interaction between climate change, digital finance, and production systems with SHAP-based interpretation.