Portrait of Deepali Kundnani

Deepali L. Kundnani

Postdoctoral Researcher, Biomedical AI

University of New Mexico Comprehensive Cancer Center · Ph.D. Bioinformatics, Georgia Tech

Working on biomedical AI

I build the systems that turn messy biomedical data into decisions, and the standards that say when to trust them.

About me

I design the methodology, not just the analysis: interpretable models over messy biomedical data, and the frameworks and evaluation standards that decide when to trust what comes out. I started at the bench, which is why I ask how data was made before I ask what it shows.

  • Now · UNM Comprehensive Cancer Center. Interpretable AI over EHR, genomic, and multi-omics data for cancer risk prediction.
  • 2025–26 · Turing. Advised and led frontier-AI evaluation: STEM expert teams built from scratch, plus the rubrics, failure-mode taxonomies, and QA systems used to judge whether a model's reasoning holds. One programme spanned 200+ subject matter experts.
  • 2019–25 · Ph.D., Georgia Tech. Mapped the roughly one million ribonucleotides embedded in every human nuclear genome and showed they are not randomly placed: the human nuclear ribome, tracking gene activity and shifting DNA supercoiling. I led the bioinformatics analyses. Published in Cell (Georgia Tech coverage); awarded the Mark Borodovsky Prize as the programme's top Ph.D. student.
  • 2015–19 · MD Anderson, Hanash Lab. Validated protein biomarkers for early cancer detection, including the panel behind this lung cancer risk model.

The thread through all of it: building the systems other people work inside, and auditing the ones that already exist, whether that is a sequencing pipeline, a clinical protocol, or an evaluation framework.

Career highlights

Research interests

Featured projects

Ribonucleotides in the human genome

Ribonucleotides in the human genome

rNMPs cluster near CpG islands and track with expression and methylation at transcription start sites.

Aicardi-Goutieres syndrome mutants

Aicardi-Goutières syndrome mutants

One mutation floods the genome with rNMPs. The other moves where they land. Only one is visible to a count.

RNASEH2A across 35+ cancers

RNASEH2A across 35+ cancers

Expression correlation with proliferation and cell-cycle markers across CCLE and TCGA.

Full portfolio

Journey

Fifteen years, four institutions, and what came out of each.

The full journey

Blog

Genomics, research practice, and the occasional detour.

All posts

Consulting

I take on a small number of outside engagements in biomedical AI.

Book a 30 minute call

Contact

Questions, collaborations, or just to say hello.

Prefer a document? View my CV, generated from this site and downloadable as a PDF.