This document presents a curated overview of my work in artificial intelligence, bringing together research projects, open-source software, technical writing, and engineering contributions developed across academia and industry. It reflects my interest in building context-aware intelligent systems that integrate perception, reasoning, retrieval, memory, and large language models to solve real-world problems.

The portfolio covers computer vision, multimodal AI, AI agents, retrieval-augmented generation (RAG), explainable AI, and context engineering, with applications in agriculture, disaster intelligence, scientific document understanding, and data analytics. Collectively, these works demonstrate a practical approach to designing reliable, scalable, and human-centered AI systems while continuously exploring emerging ideas through experimentation and technical writing.

Keywords— Artificial Intelligence, Large Language Models, Computer Vision, Multimodal AI, AI Agents, Retrieval-Augmented Generation (RAG), Explainable AI, Context Engineering.

My research focuses on the intersection of artificial intelligence research and software engineering, where I investigate how intelligent systems can be designed to address real-world challenges through the integration of perception, reasoning, retrieval, and contextual knowledge. Rather than treating individual models as complete solutions, I explore how multiple AI components can work together to create reliable, explainable, and practical intelligent systems.

Through academic research, industry experience, and open-source development, I have worked on applications spanning agriculture, disaster intelligence, scientific document understanding, and intelligent data analytics. While these domains differ, they share a common objective: building AI systems that are scalable, human-centered, and capable of supporting informed decision-making in real-world environments.

The conceptual framework presented below summarizes the research philosophy that connects the projects documented throughout this portfolio.

The research presented in this portfolio follows a project-driven methodology that combines scientific investigation with practical software engineering. Rather than focusing on isolated algorithms, each project is designed as an end-to-end intelligent system that addresses a real-world problem through the integration of modern artificial intelligence techniques.

The development process typically consists of the following stages:

1. Problem Definition — Identifying a real-world challenge and formulating a research objective.

2. Data Acquisition & Processing — Collecting, cleaning, and preparing multimodal datasets including images, text, documents, and structured information.

3. Model Development — Designing AI pipelines using computer vision, natural language processing, large language models, retrieval systems, and agent-based architectures.

4. System Integration — Combining perception, reasoning, retrieval, memory, APIs, and explainability into deployable intelligent applications.

5. Evaluation & Iteration — Assessing system performance through experimentation, qualitative analysis, and continuous refinement.

The resulting research portfolio currently consists of the following representative systems.

The diversity of these systems demonstrates a common research direction: developing reliable, explainable, and context-aware artificial intelligence capable of solving practical problems across multiple application domains.

The research methodology presented in this portfolio has resulted in a growing collection of open-source software, technical publications, and engineering contributions. Together, these outputs represent the practical outcomes of my work across academia, industry, and independent research.

This portfolio represents an ongoing journey in artificial intelligence research, bringing together academic exploration, industry experience, open-source development, and technical writing into a unified body of work. While the projects presented span diverse application domains, they share a common objective: designing intelligent systems that are reliable, explainable, context-aware, and capable of addressing practical real-world challenges.

As the field continues to evolve, my future work will focus on advancing multimodal intelligence, agentic AI, trustworthy large language model systems, and human-centered artificial intelligence through continued research, experimentation, and collaboration.

[1] P. Dave, GitHub Research Portfolio. Available: https://github.com/pooja1515

[2] P. Dave, LinkedIn Professional Profile. Available: https://linkedin.com/in/poojaddave

[3] P. Dave, Medium Technical Publications. Available: https://medium.com/@YOUR_USERNAME

[4] P. Dave, Kaggle Profile. Available: https://kaggle.com/davepooja

[5] IBM India Software Labs (ISL), Bengaluru, India.

[6] Sardar Vallabhbhai National Institute of Technology (SVNIT), Surat, India.