At the University of Florida, I am a research assistant working on AI-agent frameworks for ridesourcing forecasting and prescriptive analysis.

• Developing an AI-agent layer that integrates with spatio-temporal deep learning models (e.g., GCNs, CNN-RNN/Transformer hybrids) to enable real-time forecasting and prescriptive analytics for ride-sourcing demand.
• Designing automated mechanisms for "what-if" policy analysis, allowing simulation of interventions such as dynamic congestion pricing or infrastructure changes.
• Implementing agent-based optimization to iteratively search for interventions that optimize stakeholder objectives (e.g., minimizing passenger wait times, balancing vehicle distribution).
• Investigating scalable deployment strategies for AI agents across heterogeneous urban zones, comparing global versus local agent paradigms for improved forecasting and policy relevance.

Led research on formal methods and AI-driven feedback loops for mathematical proof automation and verification.

• Designed the FORMAL system, combining retrieval-augmented thinking with agentic feedback loops to translate natural language mathematics into Lean 4 code.
• Developed dynamic vector stores for theorem statements and tactic examples, enabling efficient proof search and verification.
• Built a robust agentic feedback loop that analyzes errors via LLMs and autonomously refines formalization processes.
• Achieved 92% syntactic correctness and 83% semantic accuracy on benchmark datasets, outperforming pure LLM approaches.
• Implemented a low-resource efficient system using local LLMs, democratizing access to formal verification.

Contributed to the development of a mental wellness app, focusing on feature engineering, AI integration, and user engagement.

• Engineered and launched multiple innovative features, including real-time push notifications and QR code-based friend addition, boosting user engagement by 30%.
• Led AI image processing setup on AWS, resolving critical issues and increasing application stability by 25%.
• Optimized front-end and back-end connectivity, preparing the app for a wider release.

Conducted research on Indigenous oral histories, environmental data, and conservation policy recommendations.

• Developed NLP pipelines to transcribe and analyze Indigenous oral histories from the Bears Ears region.
• Linked linguistic themes from transcriptions with environmental data to contextualize Indigenous narratives.
• Created a GIS-integrated storytelling platform for conservation efforts.
• Produced policy recommendations for the Bears Ears National Monument Resource Management Plan.