AI Engineer & Data Scientist · Applied AI Researcher
Scientific Analyst, Laboratory for Social Connectedness & Health, University of Arizona
I build LLM-powered systems, retrieval pipelines, and research tooling that have to survive real data — in healthcare prototypes, behavioral science workflows, and published security NLP.
Portfolio · LinkedIn · Scholar · ORCID · Email
I am a Scientific Analyst in the Laboratory for Social Connectedness & Health under David A. Sbarra. I maintain a DNA methylation bioinformatics pipeline and internal R Shiny tools for mixed-model sensitivity analysis and associational variable analysis, alongside AI methods and reproducible research workflows.
M.S. Data Science, University of Arizona · NLP Graduate Certificate · Distinguished Graduate Scholar, Spring 2026
| Project | What it is | Evidence |
|---|---|---|
| LLM-HyPZ | LLM-assisted hardware vulnerability discovery over 114,836 CVEs | IEEE ICDM 2025 Workshops, third author, DOI |
| Rural Health Kiosk / Dr. Dexter | Team graduate capstone: RAG + MedGemma + Claude for rural care access | Co-Lab Award, UArizona iShowcase Spring 2026 (team project; source not public) |
| DataLoom | Browser-native AI data pipeline studio | Live demo |
| TriageLens | Explainable ESI 1–5 triage co-pilot | Live demo · educational, not clinical |
| Systematic Review Processor | Claude PDF coder for LSCH reviews | CLI + local FastAPI UI |
| Sensitivity Analysis | R Shiny mixed-model power/sensitivity curves (simr + lme4) for clustered/twin data |
Private LSCH software |
| Associational Variable Analysis | Shiny AVA: correlation screening, incremental validity, semi-partial R², bootstrapped indirect effects | Private LSCH software |
| Methylation bioinformatics pipeline | DNA methylation processing maintained in the current LSCH role | Internal research infrastructure |
LLM-HyPZ: Hardware Vulnerability Discovery Using an LLM-Assisted Hybrid Platform for Zero-Shot Knowledge Extraction and Refinement
Yu-Zheng Lin, Sujan Ghimire, Abhiram Nandimandalam, et al. · 2025 IEEE ICDM Workshops, pp. 1268–1273
arXiv:2509.00647 · DOI
The pipeline classified hardware vs software CVEs, embedded hardware entries with text-embedding-3-large, clustered with K-means, and summarized themes. It identified 1,742 hardware-related vulnerabilities and later surfaced 411 of 1,026 CVEs used in MITRE MIHW 2025 analysis.
Multi-Class Skin Disease Classification: A Study of Transfer Learning Strategies for Deep Learning Models
IEEE IConSCEPT 2024 · DOI
- Distinguished Graduate Scholar, Spring 2026 — College of Information Science
- Co-Lab Award, Graduate Capstone — Rural Health Kiosk, iShowcase Spring 2026 (team-reported; official listing not found at rebuild time)
- Scientific Analyst, Data Science — UA Psychology / LSCH · Jun 2026–present
- Graduate Assistant, Data Science — Housing & Residential Life · Aug 2025–May 2026
- Data Analyst — Campus Health Services · May 2025–Aug 2025
- Graduate Research Assistant, AI/NLP — MITHRA Lab, College of Information Science · Jan 2025–Aug 2025
- Data Analyst — MeeBuddy · Jan 2023–Aug 2024
Self-reported analytics outcomes from public bios (15+ dashboards, 250K+ records, 85%+ forecast accuracy, and related percentages) are listed on the portfolio with attribution and still need personal confirmation.
GenAI — LLMs, RAG, prompt engineering, embeddings, clustering
ML / NLP — Python, R, evaluation, transfer learning
Data — SQL, ETL, methylation / bioinformatics pipelines, Tableau, Power BI
Product — React, TypeScript, Next.js, FastAPI, Express, R Shiny
Research engineering — reproducible pipelines, mixed-model sensitivity, AVA, documentation
I am open to AI Engineer, Applied Scientist, ML Engineer, Research Engineer, and Data Scientist roles in the United States.
Let’s collaborate: abhivn.connect@gmail.com