Comprehensive technical viva documentation featuring the system request flow, REST API endpoint schemas, Random Forest multi-class model architecture (93.4% accuracy), and production implementation code.
Official 42-page dissertation proposal detailing Random Forest recommendation modeling, DrugBank rule-based interaction screening, multi-criteria pricing decision support, and bioavailability scheduling.
Comprehensive 42-page research proposal establishing localized Sri Lankan food datasets, clinical food-drug interaction checking, ingredient-level allergy detection, and 3-tier safety risk scoring (Safe, Moderate, Dangerous).
Official 42-page proposal outlining the real-time mobile computer vision pipeline: MediaPipe Pose landmark extraction, joint angle trigonometry, state machine rep counting, and low-latency audio coaching (< 1s).
Academic 29-page proposal detailing dynamic localized user profiling via K-Means clustering, sequential LSTM neural networks for habit retention, and SHAP/LIME explainability for NCD prevention in Sri Lanka.
The definitive research dissertation for FITSHIELD, consolidating all 4 individual components, system integration algorithms, clinical validation findings, and future research trajectories.
Detailed dissertation on food-drug interaction screening rules, pharmacology CYP450 enzyme inhibition mechanisms, and nutritional contraindication mitigation strategies.
Comprehensive individual research dissertation on Random Forest multi-class optimization, biomarker-driven dosage limits, and active pharmacological contraindication filtering for personalized supplementation.
Full research findings on real-time computer vision skeleton detection, BlazePose joint angle trigonometric calculation, and real-time form correction audio cues.
In-depth study on behavioral adherence models, Markov decision processes for training difficulty recalibration, and behavioral inertia reduction loops.
All research publications and thesis dissertations undergo strict verification prior to institutional repository indexing:
Every group dissertation and individual report must satisfy the maximum 15% similarity index threshold, excluding standard references and mathematical formulations.
Submissions require formal electronic endorsement by Supervisor Dr. Chathurangika Kahandawaarachchi and Co-Supervisor Ms. Buddhima Attanayake before panel routing.
Technical notes must link to tagged Git releases with dockerized environments ensuring full algorithmic reproducibility of the ML models and CV pipelines.
The final unified thesis is permanently bound and cataloged within the SLIIT Digital Repository under Faculty of Computing Project ID R26-IT-099.
Use the standard BibTeX reference format when citing FITSHIELD technical publications or evaluation reports:
@misc{fitshield2026,
title = {FITSHIELD: Multi-Agent Intelligence for Medication-Aware Nutrition, Posture Vision, and Adaptive Habit Formation},
author = {Ramanayake, R.M.S.M. and Dilhara, W.M.P.G.S. and Herath, H.M.T.N. and Gamaachchige, C.D.},
school = {Sri Lanka Institute of Information Technology (SLIIT), Faculty of Computing},
year = {2026},
note = {Project ID: R26-IT-099}
}