Somdutta Saha, PhD, has joined UNMIRI as a Technical and Scientific Advisor. She spent the last twelve years turning genomic and clinical data into program decisions inside oncology biopharma, and that is exactly the seam UNMIRI is building on. We're glad to have her.
The work UNMIRI does lives or dies on one question: can you read an NGS report the same way across every vendor, ground the interpretation in public evidence, and defend every call back to its source. Somdutta has been on the other side of that question for most of her career, as the person who had to trust the pipeline before betting a program on it. Having that scrutiny pointed at our own architecture is worth more than another set of hands.
Who she is
Somdutta holds a PhD in Bioinformatics and Computational Biology from the University of Arkansas for Medical Sciences. Since then she's led translational bioinformatics and biomarker work across oncology and immuno-oncology at SpringWorks Therapeutics, HiFiBiO Therapeutics, EMD Serono, and GSK.
The through-line in that work is NGS-driven biomarker strategy: analyzing RNA-seq, whole-exome, and variant data to define patient populations, building the clinico-genomic analytics behind first-in-human and post-marketing programs, and assembling the data packages that go to regulators. At HiFiBiO she led the end-of-Phase-I escalation data package for regulatory submission. At SpringWorks she was the bioinformatics contact for NGS biomarker work on a clinical program. That is the kind of experience that spots where an interpretation claim is thinner than it looks, which is the most useful thing an advisor can do for a company at our stage.
What she'll work on
Her scope is deliberately narrow and technical.
- NGS interpretation architecture and methods. She reviews the technical architecture and bioinformatics methods behind UNMIRI's cross-vendor parsing and interpretation layer, the same layer covered in our cross-vendor NGS parsing post.
- The variant-identity benchmark. She helps pressure-test and, where the science supports it, expand the benchmark underneath our claim that a typed graph beats vector retrieval for variant identity. The point is to find where the claims are weaker than they look before a reviewer does. The argument itself is laid out in why vector RAG fails for oncology.
- Gold-standard validation. She validates system outputs against recognized gold standards. This is technical validation, and it is separate from clinical sign-off.
- Grant support. She is named Key Personnel on UNMIRI's NCI SBIR Phase I application, anchoring the technical-architecture and bioinformatics-methods sections. She is not the principal investigator.
A note on scope, because we try to be precise about this. Somdutta is an advisor, not an employee, and her validation work is distinct from clinical sign-off. Any work that touches real reports happens through consented, de-identified channels only, consistent with the PHI posture described in our HIPAA-ready architecture post. Her engagement is also non-exclusive and carries no conflict with her existing work, which sits outside oncology software.
Why it matters
UNMIRI is pre-revenue and building in a domain where a confident wrong answer is worse than no answer. The correction for that is people who will tell you your benchmark is soft, your validation is incomplete, or your architecture claim needs a caveat. Somdutta has spent a career being that person for drug programs. Pointing that same rigor at our interpretation layer, and at the SBIR application, is why we brought her on.
Welcome, Somdutta.
Related references
Product
NGS Interpretation API (Engine 1)
The cross-vendor interpretation layer Somdutta advises on.
Related post
Why vector RAG fails for oncology
The variant-identity argument behind the benchmark.
Related post
Cross-vendor NGS report parsing
The parsing problem the interpretation layer solves.
Reference
About UNMIRI
The team and the four-engine thesis.
Frequently asked questions
- What is Somdutta Saha's role at UNMIRI?
- Somdutta Saha, PhD, is a Technical and Scientific Advisor to UNMIRI. She advises on the technical architecture and bioinformatics methods behind the cross-vendor NGS interpretation work, helps pressure-test and expand the variant-identity benchmark, and validates system outputs against recognized gold standards. She is also named Key Personnel on UNMIRI's NCI SBIR Phase I application. She is an advisor rather than an employee, and she is not the principal investigator on the grant.
- What is her background?
- Somdutta holds a PhD in Bioinformatics and Computational Biology from the University of Arkansas for Medical Sciences and has more than a decade of translational bioinformatics experience in oncology and immuno-oncology biopharma. She has led NGS-driven biomarker and patient-stratification work at SpringWorks Therapeutics, HiFiBiO Therapeutics, EMD Serono, and GSK, spanning multi-omics analysis, clinico-genomic analytics, and regulatory data packages.
- Does this change how UNMIRI handles clinical sign-off or PHI?
- No. Advisory input on architecture, methods, and benchmarks is distinct from clinical sign-off, and UNMIRI's PHI posture is unchanged: the PHI path runs on AWS under a signed Business Associate Agreement, final clinical output is rendered by deterministic templates, and any advisory work with real reports happens through consented, de-identified channels only.
Umair Khan
Founder and CTO, UNMIRI
Building UNMIRI, a precision oncology infrastructure company with four product surfaces: cross-vendor NGS interpretation, genomics-aware decision support, oncology literature intelligence, and a free cross-vendor unification tool for clinicians. Writing here on architecture, clinical data, and HIPAA-ready AI.
Related posts
Clinical data & genomics · 8 min read
Real-Time Congress Intelligence for Medical Affairs Teams
Monitor ASCO, ESMO, AACR, ASH, and WCLC abstracts as they post using Crossref DOI metadata. How oncology medical affairs teams track congress evidence without manual triage.
Clinical data & genomics · 8 min read
Mapping Oncology KOLs From Open Data: A Practical Guide
How to identify and rank oncology key opinion leaders from public data: OpenAlex citations, ORCID and ROR IDs, congress presence, and trial PI activity. For medical affairs and MSL teams.