Loading projects...
AI-Powered Property Management & Triage System
Conversational Intake & Field Compliance
An AI-enabled property operations platform that converts resident maintenance conversations and field inspections into auditable workflows across occupancy, remediation, vendor coordination, and compliance. It combines conversational intake, QR-guided staff actions, role-based dashboards, and automated deadline monitoring in a single operating system for residential portfolios.
PythonFastAPISQLAlchemyAlembicPostgreSQLMongoDBRedisLangGraphLLM APIsAPSchedulerReportLabNext.jsReactTypeScriptTailwind CSSTanStack QueryAxios
Timeline
1 Month
Intensity
Individual
Access
Proprietary
Revision
Mar 2026
The Challenge
Defining the core problem and identified pain points that necessitated this technical intervention.
Operators managing a distributed residential portfolio were forced to run critical workflows across disconnected tools: resident complaints arrived as free-form messages, move-in and move-out checks were handled manually, inspections lacked consistent scoring and photo evidence, vendor follow-up lived in spreadsheets, and compliance escalation depended on people remembering deadlines. The business needed a unified system that could accept messy real-world input from both residents and field staff, normalize it into structured operational records, and surface live risk, ownership, and SLA status for every role in the process.
Technical Solution
The architectural and implementation strategy developed to resolve the challenge.
01
Conversational AI Intake
The backend exposes messaging webhooks that verify inbound events, batch messages per sender, enrich incoming reports with resident and property context from PostgreSQL, and invoke a LangGraph-based agent to collect issue details, identity data, images, and urgency signals — then automatically creates normalized tickets with role-based notifications.
02
Hybrid Data Architecture
A deliberate hybrid persistence model uses PostgreSQL for relational portfolio data (properties, rooms, beds, zones, residents, inspections, vendors, compliance flags), MongoDB for conversational state and AI checkpoints, and Redis for sessions and queuing.
03
QR-Guided Field Workflows
Field users are recognized by phone number and routed into deterministic mobile flows via QR tokens. Move lifecycle enforces state-machine rules: vacant-bed validation, mandatory six zero-state photos, deadline locks, and prior move-out completion requirements.
04
Inspection & Remediation Engine
Per-zone scoring with mandatory reasons for failed zones, required evidence for non-passing states, history-aware photo enforcement, and random green-score audits. Failed zones auto-open recovery cycles with repair deadlines, vendor assignment windows, and responsibility resolution driven by rotating duty schedules.
05
Automated Compliance Governance
APScheduler jobs continuously scan for overdue inspections, missed recovery SLAs, unassigned vendors, and overdue vendor completions — then deduplicate or re-notify compliance flags and auto-resolve them once the underlying issue disappears.
06
Role-Based Control Plane
A Next.js App Router dashboard with JWT-protected FastAPI calls, React Query caching, and role guards for logistics, inspection, maintenance, compliance, and admin users — plus CSV exports, inspection PDFs, printable QR forms, and admin KPI rollups.
Key Contributions
My specific roles, responsibilities, and the technical value I added to the project lifecycle.
Hybrid Data Architecture
Architected a hybrid MongoDB, PostgreSQL, and Redis persistence model that separated conversational state and AI checkpoints from relational portfolio, inspection, and compliance data.
AI Conversational Intake
Engineered an AI-driven conversational intake layer with tool-calling, per-sender message batching, resident-context enrichment, image capture, and structured ticket finalization from unstructured mobile conversations.
QR Field Workflows
Implemented QR-triggered field workflows for move-in, move-out, and inspection operations, including TTL-backed staff mobile sessions, six-photo zero-state capture, and phone-based staff recognition.
Inspection & Remediation Engine
Built a rule-based inspection and remediation engine that enforces evidence collection, auto-opens recovery cycles on failed zones, resolves responsibility through rotating bed schedules, and generates follow-up costs on persistent failures.
Compliance Automation
Automated compliance governance with scheduled deadline checks, deduplicated flag creation, acknowledgement and escalation workflows, append-only audit logging, and multi-channel alert delivery.
Full-Stack Control Plane
Delivered a role-aware Next.js control plane with cached data fetching, JWT-protected API access, evidence-rich detail views, and sanitized CSV/PDF export pipelines for operational reporting.
Performance Metrics
Results & Impact
Triage Time
< 30 min
AI conversational intake triages and structures urgent property issues in under 30 minutes — down from multi-hour delays.
Missed SLAs Reduced
45%
APScheduler compliance jobs reduced missed inspection and recovery deadlines by 45% within the first month.
Audit Pack Ready
< 15 min
Automated audit-trail generation cut hours of manual pack preparation to under 15 minutes per dispute.
Photos per Move
6
Enforced exactly 6 zero-state room photos per move-in and move-out — creating an irrefutable condition record.
Staff Roles Served
5
Single platform supports Logistics, Inspection, Maintenance, Compliance, and Admin roles with tailored dashboards.
Hybrid Architecture
3 DBs
Deliberately separated PostgreSQL (relational), MongoDB (conversational), and Redis (sessions/queues) by access pattern.