Universal AI Lab

This ambitious project seeks to streamline the ideation and execution of initiatives aimed at achieving the United Nations Sustainable Development Goals (SDGs). It envisions the use of generative AI to assist in the full project lifecycle, from conceptualization to data acquisition, analysis, and prototype development. The expected outcome is an end-to-end pipeline that enhances the ability of organizations to rapidly prototype and implement solutions, thereby accelerating progress toward the SDGs.

Goal: Create new materials to teach AI at large scale and discover the best way to learn with AI interactions. Leverage the collected data from the online students to derive meaningful insights on how to best teach this new materials.

This project involves the meticulous revision of a research paper focused on the application of machine learning to reduce air pollution. The revision process is to ensure adherence to the stringent requirements of the MSUM (Mathematical Science Publishers' Managed Submissions and Peer Review System) format. The work will enhance the clarity, structure, and presentation of the research findings, thereby facilitating the dissemination of knowledge and strategies for tackling environmental challenges through advanced computational methods.

This research project is centered on developing a comprehensive wildfire management framework. It aims to predict wildfire spread patterns by analyzing various causative factors, ranging from natural to anthropogenic. Utilizing predictive analytics and spatial data models, the project seeks to integrate meteorological data, vegetation indices, and historical fire incidents to forecast wildfire trajectories. The core objective is to enhance decision-making in resource allocation during simultaneous wildfire events. A potential outcome is the establishment of a centralized advisory system at the national level, designed to optimize fire response strategies effectively. This system would orchestrate the deployment of firefighting assets, including personnel, firetrucks, and aircraft, to extinguish fires swiftly and efficiently, thereby minimizing environmental and economic impacts.

Advise Indian government on best policy to encourage a more diverse production of cereals nationwide and advise farmers with personalized agriculture on best practices to maximize income.

This initiative proposes the creation of personalized algorithms designed to suggest minimal dietary and lifestyle adjustments with maximum impact on individual health outcomes. Employing the extensive NHANES dataset, the project will utilize advanced statistical methods and machine learning to parse through complex nutrition, diet, and lifestyle data. The outcome will be a set of actionable recommendations aimed at improving health indicators such as blood pressure, blood sugar levels, and cholesterol. The project will address the multifaceted nature of health optimization, considering individual control factors like dietary habits, alcohol consumption, and smoking.

In this project, multimodal data sources are harnessed to recommend treatment options within a clinical setting. The research combines data analytics, machine learning, and optimization techniques to analyze heterogeneous healthcare data, including electronic health records and diagnostic images. The goal is to develop a prescriptive analytics framework that can provide personalized treatment recommendations, potentially improving patient outcomes and optimizing healthcare delivery.

The research explores the symbiotic relationship between humans and artificial intelligence in the context of innovation. It examines how AI can augment human creativity to generate novel ideas and enhance the evaluative process. The project will investigate the new division of labor between AI and humans, focusing on collaborative ideation, feedback mechanisms, and the assessment of idea viability. This study aims to refine the process of ideation, potentially leading to a significant acceleration in innovation cycles across various domains.

This project focuses on applying large-language models to assist in the formulation and coding of optimization problems. By streamlining the problem-modeling process, the research aims to empower practitioners and novices alike to approach complex issues with efficient and high-quality models. The anticipated impact includes the democratization of optimization problem-solving and enhancement of decision-making quality across a broad spectrum of industries.

Numerous project ideas in mind such as:

• Explore the potential of LLM-agents to break down complex problems and build solutions to impactful problems.

• Revisit famous experiments such as the beer game and check if AI-agents reproduce human biases or not.

This research aims to explore and expand the capabilities of Large Language Models (LLMs) in reflecting the full spectrum of human response behaviors, mainly focusing on tail-end behaviors often truncated due to the centralizing tendencies of current LLMs. By examining the effects of Reinforcement Learning from Human Feedback (RLHF), prompt engineering, and decoding strategies, this study seeks to enable LLMs to simulate rare or extreme human behaviors and opinions, enhancing their usefulness in complex tasks that require creativity and precision. This project will also investigate the implications of this enhanced LLM capability in various fields, including business strategy and creative arts, and how it influences human expertise development and collaboration.

Tabular data is essential for applying machine learning tasks across various industries. However, traditional data processing methods do not fully utilize all the information available in the tables, ignoring important contextual information such as column header descriptions. In addition, pre-processing data into a tabular format can remain a labor-intensive bottleneck in model development. This work introduces TabText, a processing and feature extraction framework that extracts contextual information from tabular data structures. TabText addresses processing difficulties by converting the content into language and utilizing pre-trained large language models (LLMs).

The objective of this project is to refine and advance the codebase for an ensemble modeling system tailored for robust time-series forecasting. Utilizing adaptive robust optimization techniques, the project seeks to improve predictive accuracy and decision-making processes for a variety of applications. The research will contribute to the field by providing enhanced tools for handling uncertainty and volatility in time-series data, which is critical in sectors such as finance, logistics, and inventory management.

Goal: Given your history of conversations with ChatGPT, provide the pie chart of your usage of ChatGPT (e.g., for code, search, arts, fun, etc.)
Then, provide advice on how to improve your prompting skills. 

This innovative project aims to revolutionize museum curation through the application of generative AI. By analyzing visitor data, historical significance, and artistic themes, the project will develop a system to optimize the layout and presentation of art pieces within a museum space. The intent is to enhance the visitor experience through personalized exhibitions and to assist curators in the storytelling aspect of exhibition design. The project aspires to merge the world of art with cutting-edge technology, offering a new dimension to the appreciation and understanding of art collections. The project may involve AR/VR technology.

Explore the artistic perspective of having multi-agent generative AI systems able to design and create.

This project undertakes the challenge of forecasting federal interest rate hikes utilizing external data indicators. Through econometric modeling and sentiment analysis of financial reports and news sources, the study aims to predict monetary policy changes. The research will provide insights into the economic landscape, aiding financial institutions, businesses, and policymakers in strategic planning. The expected impact includes more informed decision-making in the face of economic shifts and potential stabilization of market reactions.

Build multimodal, spatial-temporal AI models for species distribution models (probably for birds) and prescriptive algorithms for conservation actions recommendations.