Advaith Narayana Sarva

AI and Machine Learning student at Woxsen University with international exposure from Saint Martins University. Led successful projects in AI-driven language processing and machine learning data pipelines, showcasing leadership and project management skills. Engaged in collaborative initiatives promoting knowledge sharing and community service, underscoring a commitment to innovation and social impact.

• Built a Retrieval-Augmented Generation (RAG) system to organize, retrieve, and reason over large personal and technical knowledge bases.
• Implemented semantic embeddings, contextual retrieval, and relevance scoring to improve long-term knowledge retention and recall.
• Designed the system for deterministic, explainable retrieval rather than purely generative outputs.

• Developed a Graph-based RAG architecture using ArangoDB and NetworkX to model entities, relationships, and concept hierarchies.
• Enabled multi-hop reasoning across interconnected nodes, outperforming flat vector search for complex queries.
• Improved traceability and explainability of AI-generated answers through graph traversal paths.

• Designed an NLP system to detect logical fallacies (e.g., ad hominem, strawman, false dilemma, circular reasoning) in argumentative text.
• Applied discourse analysis, linguistic pattern recognition, and reasoning heuristics to evaluate argument quality.
• Targeted applications in media analysis, debate moderation, and misinformation detection.

• Built a reasoning framework encoding mental models such as first-principles thinking, feedback loops, incentives, causality, and cognitive biases.
• Structured mental models into hierarchies / lattices to combine multiple perspectives for complex decision-making.
• Applied the framework to systems thinking, strategic analysis, and AI-assisted reasoning pipelines.

• Developed a document intelligence pipeline to extract PDF metadata, structural information, and OCR text from scanned and digital PDFs.
• Processed fields such as title, author, creation date, tags, headings, reading order, and embedded text layers.
• Improved document accessibility, searchability, and downstream NLP processing for large document collections.

• Built an NLP system that translates natural language instructions into executable CLI commands.
• Performed intent detection, argument extraction, and command synthesis for shell-based workflows.
• Focused on safe command generation, ambiguity handling, and automation of repetitive developer tasks.

• Conducted research on code-mixing and code-switching in multilingual NLP systems.
• Improved model robustness on mixed-language text (e.g., Hinglish, bilingual corpora).
• Addressed real-world challenges in multilingual social media and conversational data.

• Designed a child-friendly wearable device to enhance safety during large-scale public gatherings.
• Focused on identification, crowd-scale usability, and real-time safety considerations in high-density environments.