AppKita is a small studio of practicing physicians and software engineers building applied AI and computer-vision prototypes for medical research, education, and rehabilitation.
AppKita designs and builds web application prototypes for the healthcare sector. Our team pairs practicing medical doctors with software engineers, so every prototype is grounded in a real clinical workflow rather than a hypothetical one.
We build for medical professionals, research institutions, seminars, and students who want to explore what's possible at the intersection of medicine and applied AI — from diagnostic imaging to gait analysis to bedside risk estimation.
We currently work with select institutions and professionals through closed contracts. For collaboration inquiries, reach out via email — we respond to every message ourselves.
Built by doctors who see the workflows firsthand, from imaging review to bedside triage.
Computer vision, on-device pose estimation, and lightweight ML shipped as usable web apps.
Every prototype is designed to be tested, questioned, and iterated on — not sold as a finished product.
Each tool solves one clinical or educational problem well, instead of many problems poorly.
Twenty-seven prototypes across imaging, motion, predictive risk, monitoring, and supporting tools. Filter by domain or browse the full set.
AI-powered classifier to detect the presence or absence of intrasellar tumors from medical imaging.
Launch prototype →Multi-class classification across different intrasellar tumor subtypes.
Launch prototype →Retinopathy screening built on ConvNet.js for fast, in-browser inference.
Launch prototype →A TensorFlow.js implementation of retinopathy screening for comparison against the ConvNet build.
Launch prototype →Converts DICOM and other medical images into organized grid layouts for analysis and presentation.
Launch prototype →AI-assisted chest X-ray classification tuned for pediatric cases.
Launch prototype →Classifies MRI images into low or high Fazekas score bands.
Launch prototype →Compares a patient profile against historical ICH cases over a five-year window.
Launch prototype →Real-time body-pose detection for identifying hands-up or hands-down positions.
Launch prototype →Trains custom hands-up/down detection models on your own datasets.
Launch prototype →Interactive exergame that uses body-pose detection to encourage activity and rehabilitation.
Launch prototype →The earlier pose-controlled rehabilitation game that MoveMate 2.0 builds on.
Launch prototype →Real-time fall detection using computer vision and pose estimation.
Launch prototype →Trains and tests models for the Fall Detection prototype.
Launch prototype →An enhanced fall-detection build with improved accuracy and real-time monitoring.
Launch prototype →Training platform for the Fall Detection 2.0 system.
Launch prototype →Gait analysis with a timeline view, built for physical therapy and research use.
Launch prototype →Step detection using a smartphone mounted to a walker, for gait monitoring.
Launch prototype →A body-pose visualization wrapper that displays anatomical landmarks and movement.
Launch prototype →Estimates stroke risk from patient profile and clinical parameters.
Launch prototype →Predicts expected in-hospital duration for stroke patients from clinical parameters.
Launch prototype →Audio-based snoring detection for sleep studies and respiratory monitoring.
Launch prototype →Batch-crops images to 224×224 squares for AI model training.
Launch prototype →Text-to-speech tool for digital therapy and accessibility in healthcare settings.
Launch prototype →A quiz application covering physical medicine and rehabilitation topics.
Launch prototype →A reels recording studio with selectable webcam and mic inputs, looping video backgrounds, cameo overlays with background blur, and a presentation carousel.
Launch prototype →An ad-free Jetpack Joyride–style runner starring a tofu block as the main character.
Launch prototype →