About Me

▶ Computational Immunology & Health Equity

My dissertation, Increasing Precision and Prioritization of HLA Matching in U.S. Deceased-Donor Kidney Allocation, translates population genetics into clinical decision tools. I engineered HLA Matchability, the first U.S. individualized matching-opportunity metric (analogous to CPRA on the matching-likelihood side), validated against the national historical transplant-candidate cohort using a large National Marrow Donor Program (NMDP) nine-locus reference panel. Through the Personalized Immunologic Compatibility Likelihood (PICL) framework (frequency-weighted HLA-match priority for candidates with rare genotypes, simulated nationally on NYU’s OASIM platform), allocation can be made more equitable for Black, Asian, and Pacific Islander candidates without reducing overall transplant rates. Race is not used as a scoring variable; Matchability is applied as a continuous individual percentile.

Concurrently I built HaploFreqImpute, a high-resolution HLA imputation model using expectation-maximization over ~10M NMDP haplotype frequencies and linkage-disequilibrium patterns, to power a Probabilistic Offer Filter: a four-step simulated allocation framework calibrated on a multi-population OPTN cohort using Brier scores and the Paynter et al. open-source framework. Simulation demonstrates substantial reduction in manual histocompatibility review and expansion of the compatible donor pool. Industry partnership with Werfen (Norcross, GA) on AlleleAmbiguityUtility v2.0 for realistic SSO ambiguity simulation. The Matchability Calculator is publicly deployed via Django / AWS EC2 (transplanttoolbox.org/matchability). Two first-author and one co-authored manuscript in advisor final review for the American Journal of Transplantation.

▶ Wet / Dry-Lab Translation

Integrated bulk RNA-Seq, single-cell RNA-Seq (Seurat), ATAC-Seq, and CITE-Seq to dissect T-cell and ILC3 transcriptional, metabolic, and phenotypic programs in host-directed immunotherapy against multidrug-resistant Klebsiella pneumoniae pulmonary infection (Kolls Lab, Tulane); applied gene-set enrichment analysis and pathway analysis for mechanistic target identification, with differential-expression and signature-based biomarker discovery. Hands-on wet-lab fluency (ELISA, qPCR, flow cytometry / FACS, transgenic mouse infection models) is directly transferable to multi-omics target discovery and biomarker development across immuno-oncology, infectious disease, autoimmunity, and translational immunology.

▶ AI / ML Engineering

Built domain-specific retrieval-augmented generation (RAG) systems exposed to LLMs via Model Context Protocol (MCP) servers for citation-grounded scientific querying: HLA-RAG (~3,500 curated articles) and ILC3-RAG (~2,600). NVIDIA Deep Learning certified (PyTorch CNN / VGG16 transfer learning, >92% multi-class validation accuracy). Production use of frontier AI APIs (Claude, OpenAI, Gemini) with Cursor and GitHub Copilot for AI-augmented bioinformatics pipeline development and HPC deployment.