Product, Terra Engine & Terra Net
PropChain is a PropTech 3.0, AI-first, mobile-centric real estate platform that unifies discovery, valuation, and eventually closing into one guided workflow for buyers, sellers, homeowners, small investors, and real estate pros. Instead of static listings, PropChain adds a real-time predictive layer so users can discover, analyze, and act in one place.
Core value: "Discover, analyze, and close on property in one place—turning opaque, slow real estate workflows into a real-time, AI-guided experience for every stakeholder."
Product Roadmap & Release Plan
This roadmap is structured across four macro phases—ALPHA, BETA, GTM, and Post-GTM (Seed-close to Seed+24 months)—and is intentionally synchronized with PropChain's Seed budget envelope, data acquisition sequencing, Terra Engine training pipeline, and Terra Net scanning schedule (Phase A → A+B → A+B+C).
What is Built
Terra Engine predictive MVP, Terra Net curb-scan ingestion, mobile + dashboard experiences, MLS-backed pilots
When it's Built
24-month runway aligned with $1.4M target raise and scanning phases
Dependencies
MLS compliance, data acquisition, training pipeline, scanning schedule
Milestone Gates
Hiring, GTM spend, and reserve activation tied to validated metrics
The roadmap is written for investors, technical diligence reviewers, internal execution alignment, and ML/infra auditors from MLS partners and portfolio CTOs.
Roadmap Principles & Constraints
Scope Boundary
Included in Seed Scope
- Terra Engine predictive MVP
- Terra Net curb-scan ingestion
- Mobile + dashboard experiences
- MLS-backed pilots
Not Included
- Transaction stack
- Escrow functionality
- Marketplace features
- Funding workflows
- Multi-rig expansion >3 rigs
Runway Synchronization
All milestones must align with the 24-month runway at the $1.4M target raise. Higher-cost steps are explicitly gated by validated data, early adoption, or ARR indicators to ensure capital efficiency and risk mitigation.
1
Phase A Scanning
Initial 1-rig deployment with baseline data collection
2
Phase B Expansion
2-rig deployment gated by Phase A validation
3
Phase C Scale
3-rig deployment gated by ARR and market traction
Cost Control: Digital Marketing Lead cash activation and scanning Phase B/C are explicitly gated by validated data, early adoption signals, and revenue indicators to prevent premature capital deployment.
Data & MLS Compliance
All roadmap steps involving MLS, Trestle, ListHub, or GRID ingestion must comply with strict data governance and regulatory requirements to ensure legal operation and partner trust.
RESO 1.6+ Compliance
Full adherence to Real Estate Standards Organization data standards for interoperability
Training vs Display Restrictions
Strict separation between data used for model training and consumer-facing displays
Per-MLS Usage Restrictions
Customized compliance per MLS partner agreements and regional requirements
Data Boundary Enforcement
Pure-proprietary curb-scan data maintained separately from MLS-licensed fields
Predictive Performance Goals
Terra Engine must deliver exceptional performance to meet consumer and professional user expectations while maintaining explainability and trust.
<5s
Inference Speed
90th percentile mobile request latency target
±8-12%
Valuation Accuracy
Error band for property-level predictive valuations (non-AVM framing)
100%
Explainability
All predictions include explainability primitives for consumer and Pro audiences
Scanning Phases Drive Model Evolution
PropChain's scanning program is structured in three phases, each expanding geographic coverage and model sophistication while managing capital deployment risk.
Phase A: 1 Rig
Initial curb-scan ingestion for pilot geos. Establishes baseline data collection and validates ingestion pipeline.
Phase B: 2 Rigs
Multi-market generalization and ROI signature detection. Expands to additional states and validates cross-market model performance.
Phase C: 3 Rigs
Cross-state robustness + partner-grade datasets. Achieves enterprise-level data quality and geographic diversity.
High-Level Roadmap Overview
Phase 0 – Pre-Seed (Completed / In Progress)
For investor comprehension, not part of Seed runway. Establishes foundational technical capabilities and validates core assumptions.
Product Foundations
- Core mobile app skeleton + search surface
- CurbScan rig prototype (Jetson AGX Orin v0.7)
- Basic MLS aggregator alignment
- Early dataset mapping + raw feature inventory
Technical Validation
- Pre-Seed technical feasibility reports
- Early version of Terra Engine v0.1
- On-vehicle scanning prototype tests (5–10 miles)
- ListHub S2 Moderate Grow assessment
Phase 1 – ALPHA (Seed Month 1–6)
Primary Goals
End-to-End Predictive Loop
Deliver the first complete predictive loop integrating Terra Engine + Terra Net + mobile client
MLS-Backed Pilot
Establish first MLS-backed pilot environment with live data feeds
Proprietary Dataset
Generate minimally sufficient proprietary dataset for training v1 models
Core Features Launch
Launch safe-save ingestion, core telemetry, and initial predictive cards
ALPHA Deliverables: Terra Engine v1.0
Foundational Predictive Models
Property Embeddings
Tabular + curb-scan fusion for unified property representation
Valuation Model
Initial valuation model with non-AVM framing and transparent methodology
Market Trajectory
Trend phase estimation and early-turning-point detection algorithms
Renovation Signals
Lightweight renovation signal priors (no deep ROI model yet)
Inference Pipeline
GP/LightGBM baseline with standardized model output schema v1
Explainability
Initial explainability heuristics with feature importance and transparent ranges
ALPHA Deliverables: Terra Net v1.0
Ingestion & ETL Foundations
Core Ingestion
- GoPro → Jetson ingestion stabilized
- GPS → parcel alignment
- MLS timestamp alignment (ListHub → property)
- NAS-first → Azure ingestion bridge
Signal Extraction
- Basic curb-signal extraction (curb continuity, sidewalk condition, facade wear)
- Error retry logic (tier-1) + offline-first upload queue
ALPHA Deliverables: Mobile App
Core UX
Search & Discovery
Search + map/list UX with intuitive navigation and filtering
Predictive Cards
Predictive card v1 displaying valuation, trend, and curb-scan insights
Safe-Save Camera
Safe-save camera ingestion (capture → parcel → upload)
User Identity
Early user profile + identity model with preferences
Offline Mode
Local caching + offline read mode for uninterrupted access
Navigation Shell
Navigation shell & routing tree (Design Exhibits A–D)
ALPHA Infrastructure Requirements
NAS Environment
NAS-first environment fully configured for local data processing and storage
Azure On-Demand
Azure-on-demand configured with budgets + usage alerts to prevent cost overruns
CI/CD Pipeline
CI/CD pipeline for mobile + backend with automated testing and deployment
MLS Environments
Sandbox MLS + pilot MLS environment partitioning for compliance
ALPHA Business Deliverables
ALPHA phase focuses on validation and early feedback rather than revenue generation, establishing the foundation for future GTM activities.
1
Live MLS Pilot
At least 1 live MLS pilot in an ORI-2 market (NJ baseline) with active data feeds
2
Advisory Council
Advisory council seeded with 3–5 agents/investors providing product feedback
3
Feedback Loop
ALPHA feedback loop: weekly telemetry, model error monitoring, and iteration cycles
4
No Paid GTM
No paid GTM activities yet—focus remains on product validation and refinement
ALPHA Hiring & Spend Gates
ALPHA phase maintains strict capital discipline with minimal hiring and GTM spend, focusing resources on core product development.
Engineering Team
Cash activation for DS/MLE and FSMAD optional based on engineering throughput and velocity
GTM Spend
No GTM spend except mandatory brand/dev costs (<$5k) to maintain capital efficiency
Sales & CS
No Sales/CS hire yet—premature before product-market validation
Phase 2 – BETA (Seed Month 7–12)
Primary Goals
01
Terra Engine v2
Deliver Terra Engine v2, incorporating multi-market signals and enhanced predictive capabilities
02
Validate Terra Net
Validate Terra Net ingestion with early-scale curb data across multiple geographies
03
Pro Dashboard
Deliver Pro dashboard and full predictive cards for professional users
04
Early GTM
Begin early GTM experiments with strict KPI gates and performance monitoring
05
MLS Expansion
Launch 2 MLS pilots minimum to validate multi-market approach
BETA: Terra Engine v2.0
Multi-Geography Models
Core Enhancements
- Dual-encoder MLS + curb-feature embeddings for richer property representation
- Renovation ROI model v1 (non-AVM framing; block-level ROI signatures)
- Inference latency tuned to sub-4 seconds for optimal user experience
Expanded Predictive Cards
- Curb-score with detailed condition metrics
- Risk-score for investment decision support
- Micro-block desirability scoring
- Renovation uplift bands with ROI estimates
Performance Target: Sub-4 second inference represents a 20% improvement over ALPHA phase, critical for mobile user retention.
BETA: Terra Net v2.0
Multi-Rig, Multi-State ETL
Phase A Integration
Phase A data integrated (9,450 miles) providing baseline coverage for NJ/NY metro
Phase B Readiness
Phase B ingestion readiness (second rig) for expanded geographic coverage
Cross-Market Testing
Cross-market generalization tests: NJ → NY, NJ/NY → FL or TX to validate model portability
Expanded Curb-Signal Library
- Parking/driveway geometry analysis
- Facade wear scoring algorithms
- Vegetation overgrowth detection
- Frontage continuity assessment
- Parcel/MLS timestamp fusion v2
- Corrections for overlapping parcels
BETA: Mobile App Enhancements
Full Predictive UX
Predictive Search v2
Context-layered, trends-aware search with intelligent ranking and personalization
Draw Polygon
Draw polygon → predictive re-ranking for custom geographic searches
Saved Searches
Saved searches + recommendation triggers with automated alerts
Safe-Save v2
Safe-save ingestion v2 with batch + continuous modes for flexible data capture
Local ML-Lite
Local ML-lite heuristics for device-based ranking and reduced latency
Offline Browsing
Offline-first browsing for agents/investors with full data access
BETA: Pro Dashboard (v1)
The Pro Dashboard delivers professional-grade tools for real estate agents, investors, and property managers, transforming PropChain's predictive intelligence into actionable business insights.
Portfolio View
Comprehensive portfolio view with assets, ROI projections, and condition signals for investment tracking
Client Tracking
Client tracking + insights with engagement metrics and preference analysis
Export Capabilities
PDF/CSV exports with watermarking for professional client presentations
Briefing Mode
Predictive "briefing mode" for open houses with real-time market intelligence
BETA Business Deliverables
BETA phase introduces controlled GTM experimentation with strict performance gates, balancing growth ambitions with capital discipline.
GTM Tranche 1 Activation
GTM spend Tranche 1 ($15k) may activate if: ≥3–5 pilots active, strong NPS/product feedback, and telemetry coverage sufficient for training
Waitlist Funnel
Waitlist → early access funnel established with conversion tracking and user segmentation
Early Revenue
3–5 paying Pros (low-ACV) not required but anticipated as validation signal
BETA Hiring & Spend Gates
BETA phase begins selective team expansion while maintaining strict performance-based hiring gates to preserve runway and ensure capital efficiency.
Business Ops/PM
Business Ops/PM hire becomes justified to manage increasing operational complexity
Digital Marketing
Digital Marketing Lead remains warrant-only, no cash yet until traction validates spend
Sales/CS
Sales/CS not hired until paying seats threshold met (revenue validation required)
Phase 3 – GTM (Seed Month 13–18)
Primary Goals
1
Revenue Conversion
Convert BETA usage into subscription revenue with validated pricing and packaging
2
Predictive Performance
Ramp predictive performance using expanded curb-scan dataset across multiple markets
3
Scale Ingestion
Scale ingestion to Phase B with 2-rig deployment and multi-state coverage
4
Pro Dashboard
Deploy Pro dashboard widely to professional user base
5
GTM Spending
Begin modest GTM spending under strict KPIs and performance monitoring
GTM Technical Deliverables
Terra Engine v3.0
Robust valuation + ROI models with multi-state calibration and drift detection
MLS + Curb Fusion
MLS + curb data fusion across 2–3 states with unified feature representation
CurbScan Phase B
CurbScan Phase B ingestion (2 rigs) delivering 18,900 scanned miles
Predictive Ranking
Predictive ranking improved with drift detection and continuous calibration
UI Enhancements
Fast-follow UI enhancements: search → recommendations → alerts workflow
GTM Business Deliverables
GTM phase activates controlled marketing spend tied to validated performance metrics, ensuring efficient customer acquisition and sustainable growth.
1
GTM Tranche 2
Activation of GTM Tranche 2 ($20k) if: ≥50 SQLs, ≥10 paying customers, acceptable CAC/LTV patterns
2
Partner Content
Partner content launches (MLS, brokers, local experts) for credibility and distribution
3
Enterprise Discussions
Early enterprise/API discussions seeded with potential strategic partners
4
Predictive Insights
Predictive insights flyers for pilot MLSs demonstrating value proposition
GTM Hiring & Spend Gates
GTM phase introduces the first potential cash activation for marketing roles, but only under strict performance conditions that validate market traction and unit economics.
Digital Marketing Lead
May receive partial cash activation (≤$1.5k/mo) only if: ≥1,000 MAUs with telemetry and ≥2 pilot markets active
Sales/CS Hiring
Sales/CS still gated by revenue ($200k ACV pipeline) and retention metrics to ensure sustainable growth
Capital Discipline: All hiring remains performance-gated to preserve runway and ensure each role delivers measurable ROI before full cash activation.
Phase 4 – Post-GTM (Seed Month 19–24)
Primary Goals
01
Series A Preparation
Prepare for Series A metrics validation with comprehensive performance data
02
Dataset Expansion
Expand datasets to 3-rig ingestion (Phase C) for maximum geographic coverage
03
ARR Targets
Hit ARR targets under low/base/high scenarios demonstrating scalable revenue model
04
Compliance Hardening
Harden compliance & ML auditability for enterprise and MLS partner requirements
Post-GTM Technical Deliverables
Terra Engine v3.5
- Feature drift correction algorithms
- Model monitoring dashboards (internal)
- Calibration curves by geography
- Enhanced explainability features
Terra Net v3 & Pro Dashboard v2
- Full Terra Net v3 ingestion (3 rigs), ~47k miles
- Pro dashboard v2 with collaborative workflows
- Partner console v1 for MLS and broker integration
Post-GTM Business Deliverables
Post-GTM phase focuses on achieving Series A-ready metrics while expanding market presence and preparing for the next funding round.
$100-400K
ARR Range
Low: ~$100k | Base: ~$250k | High: ~$400k+ demonstrating scalable revenue model
2-4
New Regions
MLS partner expansion into 2–4 new regions for geographic diversification
Series A Preparation
Start packaging a WINTER 2026 raise narrative with comprehensive metrics, validated unit economics, and clear path to profitability.
Phase 5 – Seed-Complete Deliverables
Month 24 Achievements
Operational Pipeline
Fully operational Terra Engine + Terra Net predictive pipeline with proven performance
MLS Pilots
1–3 active MLS pilots in 2–3 states with validated partnerships
Product Suite
Predictive mobile + Pro dashboards released with full feature set
Curb-Scan Dataset
Full curb-scan dataset across 3 markets (~47k miles)
Revenue Generation
Revenue generation from subscriptions + early data/API partnerships
GTM Validation
GTM strategy validated for Series A with proven unit economics
Seed runway targets achieved with contingency near expected range, positioning PropChain for successful Series A raise.
Roadmap Table: Quarter-by-Quarter
Risk Map & Mitigation
MLS Approval Delays
Risk: MLS partnerships take longer than expected to approve
Mitigation: Aggregator fallback (ListHub S2), parallel MLS submission pipeline
Mitigation: Aggregator fallback (ListHub S2), parallel MLS submission pipeline
Scanning Cost Overruns
Risk: CurbScan operations exceed budget projections
Mitigation: Budget gating, reserve usage, elastic rig scheduling
Mitigation: Budget gating, reserve usage, elastic rig scheduling
Predictive Drift Across States
Risk: Models don't generalize well to new geographies
Mitigation: Drift detection dashboards, calibration per region
Mitigation: Drift detection dashboards, calibration per region
Mobile Performance Regressions
Risk: App performance degrades with feature additions
Mitigation: Offline caching + <5 sec inference requirements
Mitigation: Offline caching + <5 sec inference requirements
Over-Hiring Risk
Risk: Premature hiring burns runway without validated traction
Mitigation: Strict KPI gating for all hiring decisions
Mitigation: Strict KPI gating for all hiring decisions
Success Criteria for Seed Completion
PropChain's Seed round will be considered successful when the following comprehensive criteria are met, demonstrating product-market fit, technical excellence, and sustainable growth trajectory.
Predictive Experience Quality
Predictive experience that is fast, transparent, and trusted by agents/investors/homeowners with consistent <5 sec inference and explainable outputs
MLS Partner Validation
MLS partners agree the product enhances consumer decision-making, does not violate display/training rules, and aligns with compliance expectations
Revenue & User Metrics
PropChain reaches: $100–400k ARR, 20k+ active consumer users, 10+ Pro users across multiple markets, and complete curb-scan dataset across 2–3 states
Infrastructure & Architecture Overview
PropChain's infrastructure is engineered around cost efficiency, scalability, privacy/compliance, and predictive performance. It uses a hybrid NAS-first architecture with Azure-when-needed scaling, designed to support PropChain's Terra Engine (predictive intelligence) and Terra Net (curb-scan ingestion) while staying explicitly within the $40k cloud envelope defined in the Seed Capital Plan.
Predictive Workloads
Spanning MLS, curb-scan, and behavioral telemetry with unified feature representation
Multi-Device Ingestion
From GoPro/Jetson rigs with secure, efficient data pipelines
Mobile-First Inference
Low latency (<5 sec 90th percentile) for optimal user experience
Cross-State MLS Compliance
ListHub → RESO → MLS-direct with full regulatory adherence
Cost Containment
No uncontrolled cloud expansion or runaway storage/egress costs
Architectural Principles & Constraints
1. NAS-First, Cloud-Second
PropChain uses a NAS-first local storage stack to minimize cloud storage and egress costs. Azure is invoked only for model training at scale, selective data transformations, multi-region redundancy, and inference bursts.
This design keeps the infra envelope within ≈ $40k over 24 months, representing a 60-70% cost reduction compared to cloud-first architectures.
Cost Efficiency: NAS-first architecture eliminates the largest cloud cost drivers—storage and egress—while maintaining full scalability for compute-intensive workloads.
Deterministic Cost Boundaries
Each infrastructure component has pre-defined cost ceilings with automated enforcement to prevent budget overruns and ensure predictable capital deployment.
NAS Stack
Capped one-time + predictable monthly cost with no variable charges
Azure
Strict usage limits + enforced budget alerts with automatic shutdown triggers
MLS Ingestion
Bounded by ListHub S2 / RESO 1.6+ pricing with no surprise overages
Scanning Ingestion
Avoids cloud-heavy ingestion during Phase A/B/C to prevent runaway charges
Data-Type Segregation
PropChain enforces a strict division among data types to ensure MLS compliance, training restrictions, and explainability. This segregation is fundamental to legal operation and partner trust.
MLS-Licensed Fields
Strictly controlled with display and training restrictions per partner agreements
Proprietary Curb-Scan
Fully owned by PropChain with no usage restrictions or partner dependencies
Telemetry
User behavior and engagement data for model improvement and personalization
User-Generated Content
Safe-save camera inputs and user annotations with privacy controls
Derived/Learned Features
ML outputs and predictions with full lineage and explainability
Predictive Performance Guarantees
Infrastructure is designed to support aggressive performance targets that are critical for mobile user retention and professional user adoption.
<5s
End-to-End Inference
90th percentile latency target for complete prediction delivery
<200ms
Internal Model Inference
Core model execution time for real-time responsiveness
Supporting Infrastructure
- Efficient feature lookups with optimized indexing
- Caching of hot geographic tiles for common searches
Monitoring & Optimization
- Drift detection pipelines for model accuracy
- Continuous performance monitoring and tuning
Modularity Across States
MLS rules vary significantly by state and region. PropChain's architecture uses modular components to isolate state-specific logic and ensure compliance across diverse regulatory environments.
Modular ETL Pipelines
State-specific data transformation logic with pluggable processors
Replaceable MLS Connectors
Support for ListHub, Trestle, MLS Grid, and direct MLS connections
State-Level Policy Modules
Isolated transformations and display rules per jurisdiction
Security & Privacy First
Security is integrated—not appended later—covering the entire data lifecycle from ingestion to display. This approach ensures compliance, builds partner trust, and protects user privacy.
Encryption
Encryption at rest and in transit using industry-standard protocols (AES-256, TLS 1.3)
Access Controls
Role-based access controls per user/class with principle of least privilege
Network Isolation
Separate VNETs for ingestion vs modeling to prevent cross-contamination
MLS Display Perimeter
Isolated MLS "display-perimeter" with strict field-level controls
Audit Logs
Comprehensive audit logs for training vs display data usage with full lineage
System Overview: End-to-End
Edge Layer (On-Vehicle & Field Ingestion)
Hardware
Jetson AGX Orin rigs (Phase A/B/C)
Sensors
- GoPro 4K camera
- GPS module
- IMU (motion)
- LTE uplink
Core Functions
- Real-time recording (4K→1080p downsampling)
- GPS tagging at 1Hz for precise geolocation
- Preliminary curb-feature extraction on-device
- Local queueing when offline for reliable data capture
- Secure upload to NAS with encryption
Local NAS Layer
Primary Storage & ETL Hub
The NAS acts as the central ingestion point, providing cost-efficient local processing and storage before selective cloud synchronization.
1
Ingestion
Receives raw curb-scan video + GPS logs from field rigs
2
Partitioning
Performs partitioning by parcel ID, geo-tile, and timestamp for efficient retrieval
3
Extraction
Performs frame-level extraction & local ML-lite processing
4
Storage
Stores intermediate data for Terra Engine training with optimized formats
5
Synchronization
Syncs with Azure via scheduled or event-triggered jobs for cloud processing
Azure Cloud Layer
Secondary Processing & Model Hosting
Azure provides on-demand compute scaling and model hosting while avoiding the cost pitfalls of cloud-first architectures. Azure is not used for real-time video ingestion to prevent runaway costs.
Compute Scaling
On-demand GPU clusters during training cycles with automatic shutdown
Blob Storage
Long-term data segments with tiered storage (hot/cool/archive)
Model Training
Databricks or Azure ML for model training pipelines with experiment tracking
API Hosting
App Services / AKS for inference with auto-scaling and load balancing
Authentication
Azure AD B2C for user identity and access management
Backend API Layer
Node/Python Hybrid
Core Services
- API interface for mobile apps & dashboards
- Feature aggregation & embedding lookup
- Terra Engine inference endpoints
Supporting Services
- User authentication & authorization
- Audit + logging services
- Rate limiting and quota management
Client Layer
PropChain's client layer delivers predictive intelligence across multiple platforms, each optimized for its target audience and use case.
Mobile App (iOS/Android)
Consumer-facing discovery and analysis tools with offline-first architecture and <5 sec inference
Pro Dashboard (Web)
Professional tools for agents and investors with portfolio management and client tracking
Partner Consoles
MLS and broker tools for data integration and white-label opportunities
All clients consume standardized predictive payloads with consistent schemas and explainability metadata.
Terra Net Data Pipeline Architecture
Ingestion Pipeline
01
Edge Capture
Video + GPS/IMU → Jetson device with real-time processing
02
Edge Processing
Frame extraction, motion stabilization, rolling shutter correction, metadata tagging
03
Upload
Secure HTTPS → NAS ingestion endpoint with encryption and retry logic
04
NAS ETL
Video → frame tiles → curb-scan segments → geospatial alignment → parcel ID → MLS timestamp fusion
Terra Engine Infrastructure
Feature Library & Model Training
Feature Families
- MLS features (RESO-compliant)
- Curb-scan features (from Terra Net)
- Geospatial features (census, zoning)
- Market-level metrics (trends, inventory, DOM)
- Renovation cost priors
- User telemetry-derived features
Training Environment
- Azure ML / Databricks with GPU clusters
- Feature-store integration
- Experiment tracking with MLflow
- Model registry with versioning
- Calibration routines per geography
Model Types Supported
Gradient boosted trees (LightGBM / XGBoost), neural embeddings (dual-encoder MLS + curb-scan), temporal models (conv-temporal, LSTM-lite), and hybrid ensembles with automatic validation, drift-testing, versioning, and production promotion after QA.
Storage, Compute & Cost Governance
1
NAS Costs
One-time + monthly predictable spend supports raw storage, ETL work, hot data for training, and local compute minimization of Azure load
2
Azure Cost Controls
Strict budget alerts (daily + weekly), on-demand compute only, archival for cold data, limits on container replicas, no raw video ingestion, cross-region replication only under GTM scale milestones
3
MLS Data Cost Integration
MLS S2 Moderate Grow tier (~$20k/year) integrated into Data/Cloud $90k envelope, split between MLS data rights + transform compute, monitored monthly for overages
4
Long-Term Storage Strategy
NAS → Azure Archive as warm/cold storage with aging-based tiering; only curated training sets persist for >36 months
Telemetry & Data Capture Plan
PropChain's telemetry & data-capture strategy is built to deliver high-resolution curb-level intelligence and MLS-aligned parcel-level insights while staying within the Seed capital envelope, avoiding MLS/IDX compliance issues, and supporting scalable multi-market expansion across ORI-2 states.
This plan reflects rigorous internal engineering work, feasibility evaluations (ArchiSpect + PropPredict), and PropChain's hybrid infrastructure architecture (NAS-first + Azure-when-needed).
Objectives of the Telemetry Layer
Capture Curb-Level Intelligence
Capture curb-level street imagery, GPS traces, and environmental signals to feed Terra Net's curb-signal extraction and support predictive features
Align with MLS Data
Align field telemetry with MLS listing data to enable parcel-level matching, trend-phase detection, and curb-to-parcel fusion
Power Terra Engine
Provide stable input to Terra Engine predictive models for valuation intelligence, curb-score micro-insights, and renovation ROI estimations
Budget Compliance
Stay within strict Seed budget: Cloud $40k/24mo, Data/MLS $40k/24mo, Tools $10k, Scanning OpEx $5,900/mo per rig
MLS Compliance
Maintain MLS compliance with proprietary curb data never treated as MLS imagery, no co-mingling of restricted fields, and full separation of MLS-derived vs telemetry-derived features
Telemetry Capture Sources
1. On-Vehicle CurbScan Rigs (Primary Source)
Hardware
Jetson AGX Orin + 4K wide-FOV camera + LTE modem + SSD
Purpose
High-accuracy curb imagery + GPS + IMU + timestamping
Outputs
- Geotagged curb images (10–30 fps, compressed)
- GPS trace, heading, speed
- IMU-based camera stabilization data
- Street-level environmental features (vegetation, pavement, roof visibility)
Seed Constraint Validation
✓ Processes required frames locally on Jetson
✓ Stores imagery locally then syncs to NAS
✓ No requirement for continuous cloud uplink
✓ Stores imagery locally then syncs to NAS
✓ No requirement for continuous cloud uplink
Interim GoPro-Based Method
Seed Bridge Solution
Used in early Alpha markets or while rigs are being built to accelerate Terra Net validation and provide baseline curb-signal libraries.
Purpose & Outputs
- Rapid, low-cost street-level capture before rig deployment
- Baseline curb-signal libraries for training
- Video → processed via NAS frame slicing
- Lower-resolution but viable for training + inference prototypes
Seed Constraint Validation
✓ Very low-cost setup (~$500 vs $14k rig)
✓ Compatible with full ingestion pipeline
✓ Helps accelerate Terra Net readiness pre-rig rollout
✓ Validates end-to-end workflow early
✓ Compatible with full ingestion pipeline
✓ Helps accelerate Terra Net readiness pre-rig rollout
✓ Validates end-to-end workflow early
Mobile Application Telemetry
Ancillary Layer
Used to enhance predictive modeling and user behavior insights, providing valuable signals for personalization and model improvement.
Geolocation
Geolocation for map/search with privacy controls and user consent
Engagement Telemetry
Engagement telemetry (scrolls, taps, card interaction) for UX optimization
Geospatial Queries
Local geospatial queries & search radii for demand signal detection
Safe-Save Camera
Optional safe-save camera input (front-door, street view, property-side vantage points)
Seed Constraint Validation: ✓ Minimal network usage ✓ Local-first caching ✓ No dependency on high-volume streaming
Ingestion Workflow: End-to-End
1
Edge Pre-Processing
Frame extraction, GPS timestamping, scene validation, light curb-signal inference prep on Jetson
2
NAS Batch Sync
Nightly or base-return sync with zero cloud egress, high-throughput SSH, unified ingestion queue
3
ETL + Extraction
Road/curb segmentation, condition scoring, maintenance inference, parking/driveway extraction on NAS + Azure GPU
4
Parcel Alignment
Reverse-geocode → parcel centroids, MLS normalization (RESO 1.6+), timestamp lineage, market-phase alignment
5
Feature Store Fusion
MLS-derived + telemetry-derived features with lineage & compliance metadata for Terra Engine
Telemetry Layer Outputs
PropChain's telemetry system produces structured outputs that power Terra Engine's predictive capabilities and differentiate PropChain from MLS-only competitors.
CurbScore™
0–100 normalized scale: weighted fusion of curb condition, roof visibility & aging proxy, vegetation & facade maintenance, street-level desirability signals
Micro-Block Signature Grid
Geo-tiled 50–150m embeddings capturing noise proxies, traffic flow, street width, vegetation density, housing uniformity, price-trajectory signatures
Exterior Maintenance Index
Inferred roof/facade wear patterns based on curb imagery with confidence scores
Driveway/Parking Attributes
Binary/continuous features: driveway present, off-street parking capacity, frontage width, setback
Curb-Edge Geospatial Features
Block curvature, slope gradient, curb continuity, street quality buckets
These outputs are central to PropChain's renovation ROI model, valuation differential analysis, and trend-phase detection.
MLS Compliance Handling
Telemetry is intentionally designed to avoid common MLS compliance pitfalls through strict architectural separation and clear data lineage.
Telemetry is Never MLS-Originated
Therefore: No display restrictions, not treated as IDX media, can be shown directly to users (subject to privacy policy)
MLS-Derived Fields Remain Restricted
In feature-store schemas: /mls/* (restricted), /telemetry/* (non-restricted), /fusion/* (derived from both, with safeguards)
Training Rules Respected
MLS fields only used as permitted; non-permitted fields excluded from training sets with automated enforcement
Strict Retention Rules
Telemetry is fully owned by PropChain and long-retained; MLS data obeys per-market retention windows
Scanning Volume & Timing
Phase A/B/C Integration
1
Phase A (Months 1–9)
NJ/NY Metro (1 Rig)
9,450 scanned miles
Adequate for MVP + Alpha/Beta
Telemetry throughput fully validated
9,450 scanned miles
Adequate for MVP + Alpha/Beta
Telemetry throughput fully validated
2
Phase B (Months 10–18)
NJ/NY + FL/TX (2 Rigs)
18,900 scanned miles
Multi-market telemetry diversity
Required for Series A-level predictive performance
18,900 scanned miles
Multi-market telemetry diversity
Required for Series A-level predictive performance
3
Phase C (Months 19–24)
NJ/NY + FL/TX + CA/WA/GA (3 Rigs)
18,900 scanned miles
Multi-state calibration, cross-market drift testing, early enterprise-grade analytics
18,900 scanned miles
Multi-state calibration, cross-market drift testing, early enterprise-grade analytics
Seed-Feasible Summary: Phase A: comfortable | Phase A+B: achievable with reserve optimization | Phase A+B+C: stress-tested but feasible at $1.6M or with tradeoffs at $1.4M
Seed-Phase Infrastructure Costs
Telemetry is explicitly engineered to remain within Seed cost limits through careful architectural choices and cost optimization strategies.
Total telemetry-related spend sits entirely within planned Seed envelopes.
How Telemetry Powers Alpha → Beta → GTM
Alpha (Months 1–6)
NJ/NY curb data (GoPro + Rig v1), working CurbScore + micro-block signatures, basic valuation fusion, early mobile predictive cards
Beta (Months 6–12)
Full NJ/NY rig pipeline, FL/TX ingestion begins, parcel alignment improves → sub-30m accuracy, renovation ROI signals activated
GTM (Months 12–18)
Multi-market curb embeddings, high-confidence curb signal coverage, predictive search + ranking becomes market-ready, telemetry powers partner dashboards & pro tools
Why Telemetry Matters for PropChain
Telemetry is the differentiation engine of PropChain, providing competitive advantages that cannot be replicated through MLS data alone.
Unique Data Advantage
MLS data alone cannot capture curb conditions, micro-block desirability, exterior maintenance, or block-level ROI signatures—PropChain's proprietary telemetry fills this critical gap
Predictive Superiority
Telemetry + MLS fusion enables PropChain's predictive advantage, delivering insights competitors cannot match without similar infrastructure investment
Seed-Feasible Execution
The entire system is Seed-feasible, budget-aligned, compliant, and scalable—demonstrating technical excellence and capital discipline
This plan ensures PropChain enters GTM with an advantaged, high-density, proprietary dataset that no competitor has, positioning the company for sustainable competitive advantage and Series A success.