AI/ML for Grant Application Referral & Topic Discovery
Overview
I led and developed this AI/ML system, integrating neural networks and AI methods to support semi-automated grant application classification and scientific topic discovery.
This work was presented at Databricks Data + AI Summit 2024 and NIH AI Symposium, where I discussed AI-driven approaches for improving grant referral processes. The pilot model initially achieved 85% accuracy, and after further refinement and deployment, accuracy increased to 91%, demonstrating its real-world impact.
As part of ongoing research and experimentation, I also prototyped LLM-enhanced topic modeling and a neural network classifier with a Generative AI QA bot to explore ways AI can further improve accuracy, consistency, and decision-making in grant application classification.
*This is professional work based on publicly shared content from my Databricks Data + AI Summit 2024 presentation. The views expressed here are my own and do not represent official NIH policy.
Presented at conferences and forums including:
-
![]()
Databricks Data + AI Summit 2024
Presented technical session within the AI track of Databricks Summit
June 12, 2024
-
![]()
NIH AI Symposium 2024
Presented talk at National Institutes of Health’s AI Symposium
May 17, 2024
View the presentation video
Technical Insights: AI in Action
Key snippets from the presentation, showcasing both deployed methods and experimental AI approaches
Tech Stack & Architecture: NIH/NICHD Grant Referral System (Screenshot - Deployed System)
A look at the deployed AI system at NIH/NICHD, highlighting the neural network model and referral decision pipeline. This system assists with semi-automation of grant application referral with 91% accuracy, significantly improving efficiency and reducing manual workload
LLM-Enhanced Topic Modeling with NIH RePORTER Data (Demo Video)
This demo illustrates how LDA combined with LLMs (GPT-4) extracts meaningful research themes from NIH RePORTER data. By automating topic labeling, this approach enhances topic coherence and interpretability, improving the efficiency of grant analysis
AI-Powered Grant Referral System (Neural Network + QA Bot) (Demo Video - Experimental Prototype)
This prototype explores the potential of agentic AI in grant classification. A neural network classifier predicts grant assignments, while a Generative AI QA bot refines low-confidence cases through contextual validation. While experimental, this system demonstrates how LLMs can enhance AI decision-making pipelines