VEAI-RESEARCH
Contact Profile
PhD Research Proposal Overview

Bridging the Semantic Gap in Smart Home Monitoring:
Context-Aware Anomaly Detection and Generative AI Reporting for Parkinson's Disease

Research Objective

This research investigates GenAI-driven Ambient Intelligence and Smart Home Monitoring systems to bridge the semantic gap between clinical needs and technical implementation in Parkinson's disease care. The principal investigator is a Technical Project Manager with over 10 years of software delivery experience and 10+ years of hands-on family caregiving for a Parkinson's disease patient, currently completing an M.S. in Computer Science at Georgia Tech. This unique combination of longitudinal caregiving experience and engineering expertise positions the researcher to contribute to a Smart Environments or Assistive Technology research group by leveraging real-world longitudinal data from Japan's super-aging society context.

The Problem Statement

People with Parkinson's disease (PD) face complex, intertwined risks involving medication mismanagement, sleep disturbances, and unpredictable motor fluctuations (e.g., freezing of gait, on/off periods). Existing remote monitoring solutions are either highly intrusive (such as video cameras) or fail to interpret raw data meaningfully, presenting charts that lack actionable context for family caregivers.

The "Semantic Gap"

This research addresses the critical disconnect between raw numerical sensor outputs and the actionable, context-aware insights required by caregivers. A system is needed that preserves privacy, decodes behavioral context, and translates data into human-readable care narratives.

Core Research Questions

RQ1

Data & Ground Truth Verification

How can we reliably correlate multi-modal sensor anomalies with actual PD symptoms (e.g., freezing of gait, off-periods) using real-time, high-fidelity annotation by a resident caregiver?

RQ2

Spatio-Temporal & Contextual Reasoning

Can reasoning models reveal hidden causal relationships between environmental stressors (e.g., ambient temperature fluctuations) and subsequent daily symptom fluctuations?

RQ3

HCI & Generative AI Semantics

How can Large Language Models (LLMs) effectively bridge the "semantic gap" between raw digital signal processing and empathetic, human-readable care reports?

Proposed 3-Layered AI Architecture

A privacy-preserving, hybrid approach combining Edge Computing, Symbolic Reasoning, and Generative AI.

Layer 1 // SENSING

Privacy-First Sensing

Non-intrusive collection of telemetry via edge-computing environments:

  • Medication: Smart pillbox / NFC logging
  • Sleep & Rhythm: Under-mattress sensors & Passive Infrared (PIR) motion sensors
  • Environment: Ambient Temp/Humidity sensors
Edge-AI Enforced
Layer 2 // MODELING

Intelligent Modeling

Mapping deviations from established baseline circadian rhythms using hybrid inference models:

  • Statistical anomaly detection against personal baselines
  • Spatio-Temporal Symbolic Reasoning to identify contextual outliers
Context-Aware Core
Layer 3 // INTERACTION

Generative AI Interface

Translating mathematical and temporal data into structural clinical narratives:

GenAI Insight: "Fragmented sleep due to high room temperature may affect motor mobility today."
Semantic Interpreter

Longitudinal Methodology Plan

Phase 1 // Baseline

High-Fidelity "Ground Truth" Study (Japan Context)

Initial deployment in the researcher's home environment. Crucial Methodology Advantage: The researcher's dual role as primary resident caregiver enables real-time, high-quality ground-truth annotation of every detected sensor anomaly, generating a gold-standard baseline dataset of exceptional reliability for PD research.

Phase 2 // Scaling

Global or Regional Living Laboratory Deployment (Overseas / Japan)

Adapting and validating the architecture within broader settings, selectable between overseas environments and extended cohorts in Japan. This phase proposes converting real-world residential homes and local communities into decentralized "Living Laboratories," evaluating model generalizability and validating cross-environmental behavior recognition algorithms beyond the constraints of a traditional static laboratory setting.

Scientific

A validated framework for mapping non-intrusive sensor data directly to specific Parkinson's Disease metrics, backed by context-rich ground truth labels.

Technical

A novel architecture fusing Spatio-Temporal Symbolic Reasoning (deterministic logic) with Generative AI (semantic abstraction).

Societal

A human-centered "Care-Supportive AI" model tailored to alleviate caregiver burden in super-aging societies through real-world ecosystem scaling.

Academic Foundations & Credentials

Education Matrix

In Progress (Expected 2028)

Georgia Institute of Technology

M.S. in Computer Science (OMSCS) | Computing Systems Specialization

Graduated

University of the People

M.S. in Information Technology (MSIT)

Graduated

University of the People

Bachelor of Science in Computer Science (BS-CS)

Cloud & Methodological Validation

AWS Certified Cloud Practitioner (CLF) Active
AWS Certified AI Practitioner (AIF) Active
AWS Solutions Architect - Associate (SAA) In Progress
PMP® PMI-ACP® PSM II PSPO II ITIL® 4 MP

Anonymized Technical CV

Review full professional history stripped of personal identifiable information (PII).