I’m Arman, a machine learning researcher and software engineer working on interpretable biomedical AI—from ECG signals to drug–target affinity.
Biomedical AI Interpretable ML Research engineering
A participant-blocked permutation test asks whether 0.1 Hz amplitude modulation changes anterior–posterior sway.
Drug and protein representations interact before simulated residual quantum paths return to classical fusion.
ECG morphology is compared through angular similarity and distance from a reference.
Formal concept structure constrains an interpretable logistic risk model.
Sequence forecasts, risk-aware policy learning, and downside-risk control inform allocation.
Choose a study. Explore how the method works.
Four biomedical studies, from the research question to the evidence. Open a paper to explore its method and check a small reproduction.
All publicationsJournal of Low Frequency Noise, Vibration and Active Control · 2026
Does amplitude-modulated music change anterior–posterior standing sway once the repeated-measures design is respected and extra analyses are kept from wandering?
Scientific Reports · 2026
Can a hybrid quantum–classical multimodal model predict binding affinity when the drug, target, or both were unseen during training?
BioData Mining · 2026
Can a compact kernel combine angular similarity and distance from a normal reference to classify ECG beats with an inspectable decision geometry?
Health Informatics Journal · 2026
Can hierarchical co-occurrence structure be built into logistic regression while keeping the resulting risk model inspectable?
Molecules, ECG beats, health indicators, standing-balance trials, and market sequences call for different methods. I return to the same three questions: what structure should be represented, where could the method fail, and what evidence can a reader inspect?
From markerless posture research to Kinematum and KinemaCoach: a connected path through computer vision, model evaluation, and product engineering.
Full work experienceFeb 2026–present
Kinematum
Tehran · Full-time
Leading technical development at an AI and biomechanics company focused on markerless posture and motion analysis. Translated six months of laboratory research into a production-facing posture-analysis platform powering KinemaCoach.
KinemaCoach — AI-enabled posture assessment, motion analysis, and personalized movement guidance.
Six months developing and evaluating markerless posture estimation.
Research translated into a production-facing posture-analysis platform.
The resulting technology powers an AI-enabled movement product.
Sep 2025–Feb 2026
Neuromusculoskeletal Disorder Research Center AI Lab
Tehran · Full-time · Hybrid
Applied research on markerless human posture estimation: developing and evaluating a new computer-vision model and translating the findings into a product-ready prototype.
Follow the research into KinematumNotes on the decisions behind a model: its assumptions, its evaluation, and where it can fail.
All research notesReproducibility 3 min read
Why identity-disjoint splits, paired deltas, and an explicitly incomplete evidence grid matter more than an impressive model diagram.
Read noteBiomedical AI 3 min read
A research note on similarity geometry, patient-wise evaluation, and what the QADK study does—and does not—show.
Read noteInterpretable ML 3 min read
How structured constraints can keep a clinical-risk model inspectable, plus the validation still needed before deployment.
Read noteMy path started in computer engineering at Sharif University of Technology. Building production software taught me to care about what happens after an idea leaves the notebook.
Today, my research spans ECG signals, interpretable clinical-risk models, and drug–target affinity. I bring the same questions to each: what can we learn, how can we check it, and where does it stop working?
Have a research question or a system worth building? Let’s talk.