RESEARCHER IN BIOMEDICAL AI / MATHEMATICAL SCIENCES

Sayoni Chakraborty.

PhD Researcher · Topological Machine Learning · Biomedical Data Science

I develop mathematical and machine-learning approaches for medical imaging and spatial omics, with a focus on interpretable models, geometric and topological representations, and clinically meaningful analysis.

Explore my research View my CV
Google Scholar ↗/LinkedIn ↗/ORCID ↗
Professional portrait of Sayoni Chakraborty
SAYONI CHAKRABORTYUT DALLAS · MD ANDERSON
THE UNIVERSITY OF TEXAS AT DALLASDALLAS, TEXAS · PHD EXPECTED MAY 2027DISCOVER MORE ↓
RESEARCH DIRECTIONS

At the intersection of
mathematics & medicine.

01Topological data analysisPersistent homology · geometric structure
02Medical imaging AIClassification · segmentation · multimodal fusion
03Spatial omics & clinical dataTumor microenvironments · interpretable models
01 / BACKGROUNDABOUT · EDUCATION · PROFESSIONAL EXPERIENCE

Mathematics with clinical purpose.

I am a PhD researcher in Mathematics at The University of Texas at Dallas, working across topological data analysis, medical AI, and biomedical data science.

I design geometric and topological representations, integrate them with machine learning and deep learning, and study how they can reveal meaningful structures in medical images, spatial omics, and real-world clinical data. My research bridges mathematical methods, reproducible computation, and healthcare applications.

01

Education

EXPECTED MAY 2027

PhD in Mathematics

The University of Texas at Dallas

Research in topological machine learning, medical imaging, and biomedical data science.

DEC 2025

M.S. in Mathematics

The University of Texas at Dallas

Data Science Specialization.

MAY 2025

Graduate Certificate in Data Science

The University of Texas at Dallas

2021

M.S. in Mathematics

Jadavpur University · Kolkata, India

2019

B.S. in Mathematics

Jadavpur University · Kolkata, India

02

Professional experience

JUN 2026 – PRESENT

Research Intern / Graduate Researcher

MD Anderson Cancer Center · Houston, Texas

Joined as a research intern in summer 2026 and continued research on a reusable topological toolbox for spatial omics. Developing and evaluating methods to characterize cellular neighborhoods, tumor microenvironment organization, and tissue architecture across cancer progression stages.

AUG 2022 – PRESENT

Graduate Teaching Assistant

The University of Texas at Dallas

Leading undergraduate mathematics instruction and problem-solving sessions, supporting students across calculus, differential equations, and algebra, and grading advanced undergraduate and graduate mathematics coursework.

Teaching details ↗
RESEARCH HOME

Doctoral research conducted with the Coskunuzer Lab at UT Dallas, alongside interdisciplinary clinical and computational collaborations.

02 / RESEARCHRESEARCH DIRECTIONS · METHODS · BIOMEDICAL APPLICATIONS

Research

My research lies at the intersection of topological data analysis, machine learning, deep learning, and biomedical data science. I develop mathematically informed and interpretable learning methods for high-dimensional biomedical data, with applications in medical imaging, cancer research, spatial omics, and clinical prediction.

A central focus of my work is integrating topological and geometric representations with modern machine learning architectures to capture structural information that conventional approaches may overlook. My research spans persistent homology, multimodal feature fusion, CNNs, Vision Transformers, contrastive representation learning, and statistical modeling, with an emphasis on robust evaluation, interpretability, and real-world biomedical applications.

01

SPATIAL OMICS · CANCER RESEARCH

Topological Learning for Spatial Omics and Cancer Research

My current research focuses on developing topological and graph-based learning methods to investigate the spatial organization of cells within complex biological tissues, particularly in cancer. I am interested in how cellular heterogeneity, spatial interactions, and higher-order structural patterns within the tumor microenvironment relate to cancer development and progression.

At MD Anderson Cancer Center, I am developing a reusable topological toolbox for spatial omics analysis that integrates point-cloud representations, cell-neighborhood graphs, biologically informed filtrations, persistent homology, topology-derived descriptors, and statistical analysis. Applied to high-dimensional spatial proteomics data from lung adenocarcinoma, this work aims to characterize cellular connectivity, tissue architecture, and spatial relationships across disease stages. The broader goal is to develop interpretable, robust, and reproducible approaches for identifying biologically meaningful patterns in complex cancer datasets.

Spatial proteomicsPoint clouds & graphsPersistent homologyTumor microenvironment
Illustrative point cloud of spatially distributed cells
Illustrative cellular point cloud
ONGOING RESEARCH · MD ANDERSON
02

3D MEDICAL IMAGING · DEEP LEARNING

Topology-Infused Transformers for 3D Medical Image Analysis

My research investigates how topological representations can be integrated into deep learning architectures for analyzing high-dimensional volumetric medical images. Three-dimensional MRI and CT contain complex anatomical structures that conventional deep learning models may not explicitly capture, particularly when training data are limited.

I developed Topoformer, a topology-informed Transformer framework that incorporates sliding-band cubical filtrations and Betti-based representations into volumetric image learning. The framework integrates multiscale topological descriptors with learned imaging features and topology-aware contrastive learning to capture complementary structural information. Applications include brain MRI tumor classification and 3D medical image analysis, with an emphasis on data-efficient learning, model robustness, computational efficiency, and interpretable representations.

TopoformerSliding-band persistence3D MRI & CTTransformers
Example axial brain MRI image
Brain MRI example · Wikimedia Commons
Example abdominal CT image not from the donor liver study
CT example · Wikimedia Commons
TOPOLOGY-INFORMED DEEP LEARNING
03

MULTIPARAMETER TOPOLOGY · FEATURE FUSION

Multiparameter Persistent Homology and Contrastive Multimodal Learning

A major direction of my research involves developing hybrid deep learning frameworks that combine multiparameter persistent homology, contrastive representation learning, and multimodal feature fusion. Multiparameter persistence captures structural information across interacting filtration parameters, providing richer topological descriptors and representations of complex biomedical images.

I develop approaches that extract multiparameter topological features from dermoscopic images and integrate them with Vision Transformers through supervised contrastive learning. By fusing topology-derived descriptors with learned visual representations, this research investigates complementary structural information for skin cancer classification. The work emphasizes robust feature fusion, cross-dataset generalization, interpretable representations, and understanding the contribution of topological information to model predictions.

Multiparameter persistenceContrastive learningVision TransformersMultimodal fusion
Illustrative dermoscopic skin lesion image
Dermoscopy example · Wikimedia Commons
Illustrative histopathology slide of melanoma
Histopathology example · Wikimedia Commons
Conceptual two-parameter filtration diagram
Bifiltration concept
MULTIPARAMETER TOPOLOGICAL LEARNING
04

PERSISTENT HOMOLOGY · CANCER IMAGING

Persistent Homology and Hybrid Deep Learning for Cancer Image Classification

My research explores how persistent homology and topology-based feature extraction can reveal structural and morphological patterns in biomedical images. I develop topological descriptors using cubical filtrations, persistence diagrams, Betti curves, and vectorized persistence representations to characterize connectivity, shape, spatial organization, and multiscale patterns across image intensities.

These topology-derived features are integrated with machine learning and deep learning architectures, including CNNs and hybrid models, to complement conventional image representations. Applications include dermoscopic skin lesion classification and ovarian and breast cancer histopathology analysis. This research emphasizes interpretable feature representations, complementary structural information, and robust classification of complex biomedical imaging data.

Cubical persistencePersistence diagramsHybrid CNN modelsCancer imaging
Illustrative dermoscopic skin lesion image
Dermoscopy example · Wikimedia Commons
Illustrative histopathology slide of melanoma
Histopathology example · Wikimedia Commons
TOPOLOGICAL FEATURE ENGINEERING
05

CLINICAL PREDICTION · MULTIMODAL MACHINE LEARNING

Multimodal Clinical AI for Organ Transplantation and Predictive Modeling

My research also focuses on developing multimodal machine learning models for real-world clinical datasets, integrating medical imaging, patient-level clinical information, and topology-derived features for clinically relevant prediction tasks. I am particularly interested in high-dimensional data integration, predictive modeling, class imbalance, interpretable feature representations, and model generalization.

In collaboration with clinical researchers at the University of Arizona College of Medicine, I developed machine learning pipelines for donor liver macrosteatosis assessment using biopsy-verified CT imaging and clinical registry data from 2,709 deceased donors. This work integrates CT-derived imaging biomarkers, topological descriptors, and donor clinical variables into multimodal prediction models. My contributions include clinical data processing, topology-based feature extraction, feature engineering, model development, statistical evaluation, benchmarking, and temporal validation. The broader goal is to investigate how complementary imaging and clinical features can support organ assessment and transplantation research.

Clinical data integrationCT biomarkersMultimodal predictive MLTemporal validation
Example abdominal CT image not from the donor liver study
CT example · Wikimedia Commons
Illustration of multimodal CT topology and patient clinical data fusion
Clinical-data fusion schematic
CLINICAL AI · ORGAN TRANSPLANTATION
06

GEOMETRY & TOPOLOGY · MEDICAL SEGMENTATION

Geometry- and Topology-Guided Medical Image Segmentation

I am investigating how geometric and topological representations can complement deep learning approaches for medical image segmentation. Conventional segmentation models primarily learn pixel- or voxel-level features, while anatomical structures also exhibit important topological and geometric properties, including connectivity, boundaries, shape, and spatial relationships.

This ongoing research explores integrating topology-derived descriptors and geometry-aware representations with deep learning segmentation architectures. I am particularly interested in topology-informed feature fusion and learning strategies that can preserve meaningful anatomical structures, improve structural consistency, and support robust segmentation in complex, high-dimensional medical imaging data.

Medical image segmentationGeometry-aware learningTopological descriptorsStructural consistency
Illustrative MRI segmentation boundaries and masks
Segmentation concept
ONGOING RESEARCH · UT DALLAS
03 / PUBLICATIONSPEER-REVIEWED WORK & MANUSCRIPTS

Publications

2026 01
MANUSCRIPT · DONOR LIVER MACROSTEATOSIS

A National Biopsy-Anchored Benchmark Reveals a CT Performance Ceiling for Donor Liver Macrosteatosis Assessment

Sayoni Chakraborty et al. · Collaborative research with University of Arizona College of Medicine

Manuscript
2026 02
MIDL 2026 · PMLR 315, 2071–2096 · CO-FIRST AUTHOR

MultiPersistence Topological Fusion with Vision Transformers for Skin Cancer Detection

Fulya Tastan*, Sayoni Chakraborty*, Sangyeon Lee, Baris Coskunuzer · *Equal contribution

Publication ↗PDF ↗
2025 03
MACHINE LEARNING FOR HEALTH 2025 · PMLR 297, 23–40

Topoformer: Topology-Infused Transformers for Medical Imaging

Sayoni Chakraborty, Philmore Koung, Baris Coskunuzer · Symposium 2025; proceedings published 2026

Publication ↗PDF ↗
2025 04
ISIC WORKSHOP AT MICCAI 2025 · SPRINGER LNCS · HONORABLE MENTION

Topology-Aware Deep Models for Skin Lesion Classification

Sayoni Chakraborty*, Philmore Koung*, Baris Coskunuzer · *Equal contribution

Publisher chapter ↗DOI ↗
2024 05
TGI³ WORKSHOP AT MICCAI 2024 · SPRINGER LNCS 15239

TopOC: Topological Deep Learning for Ovarian and Breast Cancer Diagnosis

Saba Fatema, Brighton Nuwagira, Sayoni Chakraborty, Reyhan Gedik, Baris Coskunuzer

Publisher chapter ↗DOI ↗
Explore Google Scholar profile
04 / ACADEMIC ENGAGEMENTINVITED TALKS · PRESENTATIONS · ORGANIZING

Sharing research. Building community.

Invited minisymposium talks, workshop presentations, conference organization, mentorship, and academic leadership.

05 / RECOGNITION & EXPERIENCEAWARDS · SELECTIVE PROGRAMS · INDUSTRY PROJECTS

Recognition & experience.

Selected honors, supported conference travel, training programs, and collaborative applied-mathematics projects.

★ Awards & honors

  • 2026Mei-Lein Fellowship Award, UT Dallas
  • 2026Second Place Poster Award, ASA North Texas Chapter
  • 2026AIMed26 Abstract Competition — top-15 graduate student finalist (upcoming event)
  • 2025Honorable Mention, ISIC Workshop at MICCAI

↗ Travel awards & support

  • 2026SIAM Conference on Imaging Science (IS26) — Conference Travel Award
  • 2026Conference on Health, Inference, and Learning (CHIL) — Conference Travel Award
  • 2025Machine Learning for Health (ML4H) — Conference Travel Award
  • 2025DL4Sci Summer School — Travel Award, Lawrence Berkeley National Laboratory
  • 2025Mathematical Problems in Industry Workshop (MPI) — Travel Support, Claremont
  • 2024Graduate Student Mathematical Modeling Camp (GSMMC) — Travel Support, University of Delaware
  • 2024Mathematical Problems in Industry Workshop (MPI) — Travel Support, University of Vermont

◇ Selective research programs

  • 2027AxIOM: Machine Learning for Mathematics, SLMath, Berkeley — selected; scheduled
  • 2026CHIL Doctoral Symposium
  • 2026Foundations of Computational Geometry and Topology, ICERM, Brown University
  • 2025DL4Sci Summer School, Lawrence Berkeley National Laboratory

⌘ SIAM workshops & programs

  • 2025Mathematical Problems in Industry (MPI) Workshop · Claremont Colleges, California. Collaborated on RTX/Raytheon Technologies’ direction-of-arrival estimation problem, applying CosNet-based machine learning and surrogate matrix representations to signal processing. Program ↗
  • 2024Mathematical Problems in Industry (MPI) Workshop · University of Vermont. Worked with Vironix Health on analysis of chronic kidney disease data and predictive modeling of disease progression. Program ↗
  • 2024Graduate Student Mathematical Modeling Camp (GSMMC) · University of Delaware. Developed a PDE model for colonic crypt cell-population dynamics with asymmetric division, using analytical and numerical techniques. Program ↗
06 / TEACHING EXPERIENCEMATHEMATICS EDUCATION

Teaching Experience

Graduate Teaching AssistantThe University of Texas at Dallas · Department of Mathematical Sciences

Courses taught

  • MATH 2417 — Calculus I
  • MATH 2414 — Integral Calculus
  • MATH 2420 — Differential Equations with Applications

Grading experience

  • MATH 1314 — College Algebra
  • MATH 2312 — Precalculus
  • MATH 1325 — Applied Calculus I
  • MATH 1326 — Applied Calculus II
  • MATH 3379 — Complex Variables (undergraduate)
  • MATH 6303 — Theory of Complex Functions I (graduate)
  • MATH 6313 — Numerical Analysis (graduate)

Led calculus and differential equations problem-solving sessions, provided instructional support, and assessed undergraduate and graduate mathematics coursework.

Private Mathematics & Science TutorIndependent tutoring · High-school students

Provided one-on-one tutoring in mathematics and science to high-school students, with primary emphasis on mathematics.

07 / DOCUMENTSFOR RECRUITERS & ACADEMIC COLLABORATORS

The right level of detail.

A concise, one-page overview for hiring teams and a more complete two-page CV for research and academic opportunities.

ONE PAGE · RESUME

Professional résumé

Research experience, selected publications, technical skills, and measurable impact — optimized for a quick review.

Download résumé
TWO PAGES · CV

Research & academic CV

Expanded research profile including publications, presentations, awards, selective programs, teaching, and leadership.

Download CV

08 / CONTACT

Let's connect across disciplines.

I welcome research collaborations and conversations about biomedical AI, scientific machine learning, and academic or industry research opportunities.

LOCATIONDallas, Texas, United States
FIND ME ONLINEGoogle Scholar ↗LinkedIn ↗ORCID ↗
For my CV and résumé, visit the separate documents section ↗.