I built CuraLink, an agentic AI platform that helps biomedical researchers explore literature, retrieve up‑to‑date evidence, and generate structured research reports while never acting as a diagnostic or treatment tool.
I adopted a multi‑agent research workflow that pipelines research planning, evidence retrieval, RAG, PICO extraction, evidence comparison, LLM synthesis, citation verification, and report generation.
I integrated real‑time biomedical retrieval from PubMed, OpenAlex, and ClinicalTrials.gov so the system always works with the latest publications.
I built a document‑based RAG pipeline that parses PDF and TXT uploads through OCR, layout detection, text extraction, chunking, and metadata tagging before indexing for retrieval.
I created a dynamic medical knowledge graph that stores entities such as genes, diseases, drugs, and trials and enables multi‑hop traversal to discover related or conflicting findings.
I added citation verification that cross‑checks extracted identifiers (PMID, DOI, NCT) against retrieved documents and flags mismatches.
I implemented security guardrails including prompt‑injection scanning, PII/PHI redaction, and workspace‑level isolation for each user.
I added token, inference‑cost, latency, and estimated‑savings tracking in a persistent usage ledger for observability and expense management.