Research Fellow
Harvard Medical School
Department of Neurology
PHYSICIAN-SCIENTIST · AI ENGINEER
Alon
Gorenshtein, MD
Building human-centered AI for neurology — across neurocritical care, neural interfaces, and the systems that shape clinical decisions.
About
Medicine meets
machine intelligence.
Alon Gorenshtein, MD, is a physician-scientist and AI engineer affiliated with Beth Israel Deaconess Medical Center and Harvard Medical School. He is Head of AI Engineering at BRIDGE GenAI Lab and a postdoctoral fellow in the BIDMC Epilepsy + Data Science Lab. His research spans clinical AI safety and evaluation, neurophysiology, brain–computer interfaces, and neurocritical care.
My work spans clinical LLM evaluation, epilepsy data science, brain–computer interfaces, and the design of tools that can make complex clinical workflows safer and more useful. I care about the moment when a promising model meets a real patient, a real clinician, and a real decision.
Today I’m a Research Fellow at Harvard Medical School and a Postdoctoral Research Fellow in the Epilepsy Division at Beth Israel Deaconess Medical Center. I also lead AI Engineering at BRIDGE GenAI Lab.
Professional profile: Harvard Medical Faculty Physicians on BRIDGE GenAI Lab.
Visit BRIDGE GenAI LabPostdoctoral Fellow
BIDMC Epilepsy + Data Science Lab
Department of Neurology
Head of AI Engineering
BRIDGE GenAI Lab
Safe AI. Real Medicine.
Research focus
Where intelligence
meets responsibility.
My work asks how intelligent systems can remain useful and answerable to people: in high-stakes clinical care, through neural interfaces, and in the objectives we give our systems.
The connecting question
How do we build systems that extend human capability without losing human judgment?
AI misalignment
Studying the gap between what an AI system optimizes and what people actually need: context, judgment, empathy, and responsibility.
Generative AI in BCI
Language models and neural interfaces for communication, control, and restoring agency in neurological disease.
Neurocritical care & consciousness
Neurocritical care and disorders of consciousness, where clinical examination, recovery, and uncertainty meet at the bedside.
Research in depth
Questions, methods,
and evidence.
Clinical AI safety and evaluation
How clinical AI is tested, where benchmark results mislead, and what structured neurological reasoning can establish.
Read the research →AI in clinical neurophysiology
INSPIRE, nerve conduction studies, EMG reporting, and the evidence from a randomized trial of physician–AI collaboration.
Read the research →Language models and brain–computer interfaces
A systematic review of how language models support communication and control, and where clinical evidence is still missing.
Read the research →Neurocritical care and consciousness
A cohort study of when command-following returns after acute brain injury, how it fluctuates, and the role of sedation.
Read the research →Research projects
Ideas made
into tools.
Research systems and benchmarks built with collaborators to investigate clinical AI, neurology, and evidence synthesis.
RESEARCH PROJECT01 / 04
INSPIRE
Alon Gorenshtein and collaborators developed INSPIRE, an agent-guided research system for interpreting and reporting nerve conduction studies and electromyography.
Explore the projectRESEARCH PROJECT02 / 04
LITERAS
LITERAS is a multi-agent biomedical literature retrieval and review framework developed by Alon Gorenshtein and colleagues using PubMed-based search and citation validation.
Explore the projectRESEARCH PROJECT03 / 04
LitBench
LitBench is a research benchmark by Alon Gorenshtein and colleagues for retrieval-grounded, multi-paper evidence synthesis in epilepsy.
Explore the projectRESEARCH PROJECT04 / 04
Neuro-Copilot
Neuro-Copilot is a clinical AI research framework studied by Alon Gorenshtein and colleagues for neurological decision support in the emergency department.
Explore the projectA COLLABORATIVE HOME FOR CLINICAL AI
Building better AI,
together.
At BRIDGE GenAI Lab, we study the safety, alignment, and real-world behavior of clinical AI. I lead AI engineering alongside a team of physicians and researchers.
Explore BRIDGE GenAI LabFOLLOW THE WORK
Selected work
Work grounded in
three questions.
Selected studies on clinical uncertainty after acute brain injury, generative AI in neural interfaces, and evaluation beyond a narrow task objective.
AI misalignment
When systems meet human context
Evaluating whether clinical AI remains aligned with patient needs, clinician judgment, and the responsibilities that matter beyond a narrow task objective.
Read the paperGenerative AI in BCI
Language models for communication and control
A systematic review of how large language models are being integrated into brain–computer interfaces for assistive communication and control.
PubMed recordNeurocritical care
The unsteady return of command-following
A large-cohort study showing how bedside command-following after acute brain injury can return early and fluctuate, with sedation providing important context for interpretation.
Neurocritical CarePath to here
A nonlinear
clinical path.
Medicine, engineering, and research have never been separate tracks. They are the same work viewed from different angles.
Harvard Medical School · BIDMC
Research Fellow and Postdoctoral Research Fellow in Neurology, Epilepsy Division. Head of AI Engineering at BRIDGE GenAI Lab.
Icahn School of Medicine at Mount Sinai
Research Fellow in the Windreich Department of AI & Human Health, developing AI-agent methods for clinical workflows.
Rambam Health Care Campus · AI in Neurology Lab
Researcher in the AI in Neurology Lab, building INSPIRE and Neuro-Copilot across clinical neurophysiology and decision support.
Bar-Ilan University · Azrieli Faculty of Medicine
Doctor of Medicine, with research across neurology, neuroimmunology, clinical AI, and medical evidence synthesis.
University of Debrecen · Faculty of Medicine
Preclinical medicine and teaching assistant in neuroanatomy, leading cadaveric laboratories and tutorials.
Selected publications
The work,
in the literature.
Peer-reviewed research, accepted manuscripts, and ongoing work across clinical AI and neurology.
Google ScholarNature Medicine · 2026How to meaningfully evaluate AI in clinical medicinemedRxiv · 2025 · Preprint versionAI Agents in Clinical Medicine: A Systematic ReviewBiomedical Physics & Engineering Express · 2026Large language models integrated into brain–computer interfaces for communication and controlClinical Neurophysiology · 2025Agent-guided AI-powered interpretation and reporting of nerve conduction studies and EMGPLOS Digital Health · 2025A multi-agent approach to neurological clinical reasoningJournal of Neurology · 2025AI-based EMG reporting: a randomized controlled trial
Browse selected publications and download citations →Research software and public code →
LET’S TALKGood work is
Good work is
collaborative.
Research partnerships, clinical AI, and neurotechnology with a human point of view.
OPEN TO CONVERSATIONS ABOUT
Hard problems in clinical AI, neurology, and neurotechnology. If the question matters to patients or the people caring for them, I’d be glad to hear from you.
Email Alon agorensh@bidmc.harvard.edu