case study EXEGESIS

Exegesis — Academic Research Portal & RAG AI

• Academic institutions lack a centralized system to manage faculty profiles, research output, and scholarly metadata in one place.
• Research assistance is manual — answering questions about academic profiles, publications, and competency metrics requires digging through disconnected databases.
• Author disambiguation is hard — matching publications to the correct author among multiple candidates with similar names is error-prone and time-consuming.
• External scholarly data is fragmented across OpenAlex, Crossref, and DataCite with no unified search or normalization layer.
• Institutions need AI tools that are grounded in their own data, not generic LLMs that hallucinate academic credentials.

• Built a full RAG AI pipeline — input classification, context assembly from 6 sources (profiles, research catalogs, scholarly metadata, competency analyzer, faculty database), and Ollama LLM inference with SSE streaming.
• Designed a Groq AI disambiguation system that ranks author candidates by likelihood, enriches publications with Scopus indexing status and AI-generated match evidence, and computes level recommendations on a 1-5 scale.
• Implemented a multi-provider publication search (OpenAlex, Crossref, DataCite) with DOI fallback chain and unified normalization through a SourceAdapter layer.
• Created an Academic Competency Index that quantifies research output, publication quality, and scholarly impact through a transparent methodology.
• Added AI guardrails with topic classification, prompt injection detection, scope enforcement, and PII filtering across all AI interactions.