Aryan
Singh Dalal

Combined large language models, retrieval-augmented generation, natural language inference, and ontology learning in Echo-LLM and PERSEUS to produce evidence-grounded knowledge and reduce unsupported or hallucinated relations.

Designed workflows that turn text, image, and audio evidence into knowledge graph and ontology structures with provenance. Automated verification checks each result, and people review it before use.

Built WiseOWL, which scores ontologies on documentation coverage, label-definition alignment, structural connectivity, and hierarchical balance.

Applied machine learning and computer vision to agricultural and thermal/space systems, including sprayer-boom displacement measurement and condensation and freezing analysis from microscopy images.

Project manager on SPAR-KG, a multi-investigator DoD-funded project with technical work across computer science, engineering, and domain teams. General co-chair and proceedings chair of the international LLMS4KGOE 2026 workshop at ESWC 2026.

PhD research: PERSEUS (Perceptual Semantic Extraction & Unified System), a trustworthy multimodal knowledge-graph construction and ontology-learning framework using LLMs, hybrid retrieval, natural-language-inference verification, provenance-aware grounding, and human-in-the-loop validation. Major advisor: Dr. Hande Küçük McGinty.

Thesis: Quantifying Target Movement Using Neural Networks. Major advisors: Dr. Ajay Sharda and Dr. Pascal Hitzler.

Developed and evaluated LLM-assisted and neuro-symbolic workflows for ontology learning and knowledge graph construction, including retrieval, semantic validation, evidence-aware verification, provenance, ontology reuse and quality, multimodal knowledge graphs, hallucination mitigation, and human-in-the-loop validation. Managed interdisciplinary research spanning knowledge engineering, scientific AI, agriculture, food systems, and space-related applications.

Developed computer vision and machine learning methods for agricultural systems, including automated quantification of sprayer boom displacement, and supported grant writing and interdisciplinary work involving agricultural machinery, sensing, computer vision, and applied AI.

Completed an applied machine learning project involving image-based prediction and COVID-19 safety and precaution monitoring.

Multimodal knowledge graph construction pipeline with retrieval, NLI verification, and provenance grounding. Python.

Ontology descriptiveness and semantic correctness scorer. Python.

Re-ranks BioPortal ontologies by topical relevance and ontology quality to support reuse.

LLM-assisted ontology learning workflow with Neo4j deployment.

Eight-week online undergraduate course with 12 students. Built all assessment materials, including a 228-question quiz bank and the midterm exam; also holds office hours and grades.

Held office hours for a graduate course on knowledge graphs and semantic technologies.

Gave four guest lectures on heaps and sorting to six students, one of them online. Designed a fifteen-question quiz that gave credit for valid approaches outside the listed answer options.

Graded assignments, held office hours, and proctored the midterm and final project presentations for about 60 students.

Mentored undergraduate teams during a weekend-long national data analysis competition. Guided teams through exploratory analysis and modeling strategy, and helped them turn their findings into five-minute judged presentations.

Trained a master's student in image annotation, annotation quality standards, and computer vision basics. Her annotations became part of the 1,300+ image microscopy dataset used in the project's paper in Frontiers in Thermal Engineering (2026).

Trained and supervised a five-person annotation team of graduate students and visiting scholars. Defined the bounding-box labeling protocol, assigned 700 images to each member, reviewed every member's annotations, and gave individual corrective feedback. The team's annotations formed the base dataset, later augmented to 9,485 training images, for the model published in Computers and Electronics in Agriculture (2026).