2025
BaMANI: Bayesian Multi-Algorithm Causal Network Inference using Machine Learning and Deep Learning Frameworks ↗
arXiv preprint · open ensemble pipeline for causal network inference from high-dimensional biomedical data
First author
AI & Data Science for Medicine
Habib LatifizadehMachine learning · Causal inference · Computational biology
Trustworthy AI and rigorous statistical modeling for multimodal biomedical and physiologic data — built for translational clinical impact.
I develop machine-learning, causal-inference, and multi-scale modeling methods that turn complex clinical, physiologic, and molecular data into reliable, uncertainty-aware evidence for medical decision-making. My work sits at the interface of data science, biomedicine, and computational biology.
NowProject Co-Investigator & Visiting Scholar, MIT Laboratory for Computational Physiology (MIT Critical Data) — 2026–2027
2024–2026Postdoctoral Associate, Department of Biostatistics & Bioinformatics, Duke University
FundedNamed Personnel, NIH/NIAID R01 AI173333 — Duke–Weill Cornell translational immunology project
Research Bio
Clinical AI is accurate on average and unreliable at the edges that matter — the individual patients, subgroups, and physiologic moments where decisions are hardest. My work pairs causal reasoning, mechanism, and calibrated uncertainty so models are honest about what they don't know.
As Project Co-Investigator and Visiting Scholar at the MIT Laboratory for Computational Physiology (MIT Critical Data), I lead reliability and harm evaluation of multimodal clinical AI across imaging and electronic health records. My work sits at the interface of machine learning, causal inference, computational biology, and translational clinical research — turning high-dimensional biomedical, physiologic, and molecular data into reliable, uncertainty-aware evidence, spanning causal Bayesian networks, generative modeling, optimal transport, physics-informed neural networks, and multi-scale mechanistic modeling.
As a Postdoctoral Associate in the Department of Biostatistics & Bioinformatics at Duke University, I contribute to an NIH-funded Duke–Weill Cornell Medicine project (NIH/NIAID R01 AI173333, Permar Lab) on immune correlates of protection, building multi-scale mechanistic and statistical-learning models. I hold a Ph.D. in Applied Mathematics & Data Science.
This trajectory — from a causal-network dissertation and deep-learning work in computational biology at WVU, through an NSF Digital Health fellowship, to multi-scale modeling at Duke and clinical-AI reliability at MIT — is a deliberately interdisciplinary one, aimed at the collaborative space between medicine, statistics, engineering, and the computational sciences. Across it I emphasize reproducible pipelines and explicit uncertainty quantification as part of the modeling task, not afterthoughts.
Featured
Active · 2026–2027
MIT Critical Data
Project Co-Investigator and Visiting Scholar at the MIT Laboratory for Computational Physiology — reliability and harm evaluation of multimodal clinical AI.
Research experience → Nature Communications · 2022Heterocellular networks in cancer
Co-authored study on how oncogenic gene expression locally reshapes cell-to-cell regulatory networks and the tumor immune microenvironment.
Read publications → Open Source · arXiv 2025BaMANI causal inference
Bayesian multi-algorithm causal-network inference — ensemble structure learning with calibrated, bootstrap-stable uncertainty on every edge.
Explore research →Research Vision
How do we build AI for medicine that clinicians and scientists can trust — because it respects the underlying biology and physiology, and reports what it does not know?
My independent program pairs mechanism-informed modeling, causal reasoning, and calibrated uncertainty across three directions spanning physiologic, clinical, and molecular data — each grounded in completed methodological work and sized for independent extramural funding (NIGMS, NHLBI, NIBIB, NICHD).
D1
Trustworthy Multimodal Clinical AI
Reliability and harm evaluation of models over imaging and electronic health records — calibration, conformal uncertainty, and subgroup/fairness auditing for safe clinical decision support.
D2Causal Inference for Patient Selection & Outcomes
Treatment-effect heterogeneity and reproducible phenotyping from observational clinical data — asking for which patients an intervention helps, and why.
D3Multi-Scale Modeling & Computational Biology
Mechanistic simulators coupled with scientific ML for immunology and translational medicine — physics-informed surrogates and network biology across molecular to physiologic scales.
Research
Mechanism-informed, uncertainty-aware AI for biomedical and clinical data.
A single question organizes the program: how do we build models for clinical and biomedical decisions that are accurate at the edges that matter most — and that report what they do not know? The work spans four themes with a shared core in causal reasoning, mechanism, and calibrated uncertainty.
- Theme 01
Trustworthy & Reliable Clinical AI
Calibration, conformal prediction, consensus harm-scoring, and subgroup & fairness auditing for perioperative and multimodal clinical AI.
- Theme 02
Causal Inference & Bayesian Networks
Ensemble Bayesian structure learning (BaMANI) with MCMC, conditional-independence testing, and bootstrap stability.
- Theme 03
Multiscale Scientific ML
Mechanistic simulators coupled with GP surrogates and physics-informed neural networks for translational medicine.
- Theme 04
Reproducible Pipelines & Harmonization
Optimal transport with relaxed marginals, graph clustering, and containerized HPC for cross-dataset harmonization.
Research Experience
Research trajectory across scientific problems and data modalities.
- MIT · 2026–2027
Trustworthy Clinical AI
Reliability & harm evaluation of multimodal clinical AI — deep-learning ensembles, consensus harm scoring, and subgroup auditing across MIMIC-IV / MIMIC-CXR.
- Duke Epilepsy Center · 2026–2027
Surgical Decision Support
Lead Data Scientist & Biostatistician on SEEG 5-SENSE prediction of postoperative seizure freedom — ROC/AUC, calibration, and causal subgroup analysis across resection, LITT & neuromodulation.
- Duke · 2024–2026
Multiscale Scientific ML
Mechanistic simulators coupled with GP surrogates and physics-informed neural nets; cross-dataset harmonization via optimal transport (NIH R01 congenital-CMV).
- NSF Fellowship · 2023–2024
Causal Network Inference
Scalable ensemble Bayesian structure learning with bootstrap stability and interpretable ML — the foundation of BaMANI.
- ML Research · WVU Cancer Institute · 2020–2023
Generative & Causal ML
VAEs, WGAN-GP, and MCMC-based causal discovery for high-dimensional biology (Nature Communications, 2022).
Selected Publications
Current-field work in causal inference, biological data science & scientific ML.
2022
Data-driven learning how oncogenic gene expression locally alters heterocellular networks ↗
Nature Communications, 13(1): 1986
Co-author · method development and computational analysis
Press coverage (2022): AAAS / Science-affiliated press, TechiLive, and WVU Research & Graduate Education.
Ongoing Projects
Current research & collaborations.
Active · 2026–2027
Reliability & Harm Evaluation of Multimodal Clinical AI
At the MIT Laboratory for Computational Physiology, evaluating ensembles of deep-learning diagnostic-imaging models via a consensus-based harm score, and linking image-level error to ICU care phenotypes across MIMIC-IV and MIMIC-CXR with calibration, uncertainty quantification, and subgroup/fairness auditing.
Research experience →Active · 2026–2027
Surgical Decision Support for Epilepsy Outcomes
As Lead Data Scientist & Biostatistician at the Duke Comprehensive Epilepsy Center, leading the statistical design and reproducible analysis of a temporal-lobe-epilepsy study evaluating whether the SEEG-derived 5-SENSE focality score predicts postoperative seizure freedom — assessing discrimination, ROC/AUC with bootstrap CIs, calibration, and clinically meaningful thresholds across resection, LITT, and neuromodulation, with causal-explanatory subgroup and failure analysis.
Research experience →Active · NIH-funded
Immune Correlates of Protection in Congenital CMV
Named Personnel on NIH/NIAID R01 AI173333, leading in-silico vaccine-efficacy modeling and maternal–fetal immune analytics through a multi-scale scientific machine-learning framework — a joint Duke–Weill Cornell Medicine collaboration.
NIH RePORTER →In preparation
Cross-Species Cell-Type Mapping via Optimal Transport
Lead author on a methods manuscript using relaxed-marginal probabilistic optimal transport for reproducible cross-species alignment of flow-cytometry measurements — supporting standardized biomedical data integration and transferable out-of-sample inference.
Duke CHSI →Preprint · arXiv
BaMANI — Causal Network Inference Toolkit
Combines constraint- and score-based structure learning with conditional-independence testing, multiple-hypothesis correction, and bootstrap stability.
arXiv preprint →Teaching & Mentorship
Methods as a craft — derivation, code, and a worked example.
I treat calibration, subgroup performance, and uncertainty as part of the modeling task itself, and design every course around a reproducible artifact students can rerun. Five graduate courses drawn from my research program, from statistical learning for health data to uncertainty-aware clinical ML.
Teaching philosophy, syllabi & mentoring →Awards, Fellowships & Funding
Selected recognition & funding.
2025
Excellent Editor AwardIJMSSC · World Scientific
Service & Academic Leadership
Editorial boards & peer review.
View full service profile →
Selected Talks
Invited & conference presentations.
2026
Cross-Species Immune-Cell Mapping via OT-RMC for CMV Vaccine Research
Invited Speaker · Duke CHSI / MISM Center of Excellence
2023
Data-Driven Modeling for Inferring Key Protein Impact on Cell-to-Cell Networks
Invited Speaker · BridgesDigital Health NRT, WVU
Get in touch
Let's talk.
Open to collaboration, student inquiries, and academic recruiting conversations — especially with colleagues working in clinical AI, computational physiology, causal inference, and computational immunology.