Research Canvas: A Privacy-Preserving, RAG-Powered Intelligent Research Assistant Platform Built on Local LLM Inference and Microservice Architecture

     Research Canvas presents a comprehensive, privacy-preserving platform for AI-assisted research that addresses the converging challenges of information overload, data privacy, and the need for source-grounded generation in scholarly work. By integrating local LLM inference via Ollama/Gemma4, Retrieval-Augmented Generation, Named Entity Recognition, and a modular microservice architecture, the system supports the full research lifecycle within an isolated, user-controlled environment. The canvas-based workspace design facilitates the nonlinear, iterative nature of research ideation, while asynchronous processing enables computationally intensive tasks—from corpus-wide NER to automated literature review drafting—without disrupting interactive workflows. The platform builds upon and synthesizes advances across multiple active research fronts: RAG-based domain-specific assistants, neural NER architectures, AI-assisted academic writing, microservice NLP pipelines, and structured canvas-based design methodologies. While the current work is primarily architectural, the design decisions are grounded in empirically validated principles from the literature. Future work prioritizing formal evaluation, hybrid inference, and collaborative features will be essential to realizing the platform’s full potential as a transformative tool for the research community.

1.      L. Kristoffersson, “Evaluating Techniques for Building AI Assistants in a Specialised Domain: Comparing RAG Techniques and LLMs in Custom Built Assistant for the Canvas LMS,” 2024. Available: https://www.diva-portal.org/smash/record.jsf?pid=diva2:1897037 2.      Z. Hutchens and Q. Cheng, “WIP: Automated Extraction of Domain-Specific Concept Maps Using a RAG-LLM Ensemble Framework,” in IEEE Frontiers in Education Conference, 2025. Available: https://ieeexplore.ieee.org/abstract/document/11328309/ 3.      D. J. Liebling, M. Kane, M. Grunde-McLaughlin, I. J. Lang, S. Venugopalan, and M. P. Brenner, “Towards AI-assisted Academic Writing,” arXiv preprint arXiv:2503.13771, 2025 4.      S. R. Chavan, “Optimizing Web Design Code Generation: A Comparative Study of Finetuning, Pretrained Models, and RAG,” San Jose State University, 2024. Available: https://scholarworks.sjsu.edu/etd_projects/1456/ 5.      K. Pu, K. J. K. Feng, T. Grossman, T. Hope, B. D. Mishra, M. Latzke, J. Bragg, J. C. Chang, and P. Siangliulue, “IdeaSynth: Iterative Research Idea Development Through Evolving and Composing Idea Facets with Literature-Grounded Feedback,” arXiv preprint arXiv:2410.04025, 2024. 6.      G. van Capelleveen, C. Amrit, D. M. Yazan, and H. Zijm, “The recommender canvas: A model for developing and documenting recommender system design,” Expert Systems with Applications, vol. 129, pp. 14–29, 2019. 7.      B. M. Napoleão, F. Petrillo, and S. Hallé, “A Tertiary and Secondary Study Canvas,” arXiv preprint arXiv:2004.01315, 2020. 8.      A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention Is All You Need,” arXiv preprint arXiv:1706.03762, 2017. 9.      G. Lample, M. Ballesteros, S. Subramanian, K. Kawakami, and C. Dyer, “Neural Architectures for Named Entity Recognition,” arXiv preprint arXiv:1603.01360, 2016. 10.   P. Liu, Y. Guo, F. Wang, and G. Li, “Chinese named entity recognition: The state of the art,” Neurocomputing, vol. 473, pp. 37–53, 2022.