Inspiration & The Challenge Land development for renewable energy projects (solar, wind, battery storage, and hydro) involves navigating complex, fragmented legal contracts and strict environmental regulations.
Without an automated system, compliance management is severely inefficient.
Disconnected Stakeholders: SCADA engineers often lack visibility into legal obligations tied to specific land parcels, while Land Agents rely on disconnected spreadsheets, emails, and notes without real-time tracking.
Manual Compliance Monitoring: Environmental conditions, lease restrictions, and operational activities must be monitored across numerous properties, creating significant administrative overhead and increasing the risk of overlooked violations.
What parcelsite Does Parcelsite transforms manual lease compliance into a real-time, automated workflow:
Automated Lease Extraction (NLP): Users upload lease contracts or PDFs; our NLP engine automatically parses boundaries, operational restrictions, time-limited clauses, and environmental obligations into a structured schema linked to a specific Parcel ID.
Geospatial Land Mapping (GIS): Renders parcel boundaries via GeoJSON/KML uploads, creating visual layers for buffer zones, protected land-use areas, and buildable regions. Lease rules become spatially anchored directly to the land.
Virtual Sensor Ingestion: Continuously streams live weather, air quality, and soil telemetry, timestamping and geo-matching data to every parcel.
Automated Violation & Change Detection: Automatically flags rule conflicts (e.g., work occurring above a 35°C threshold) with live alerts, visual logs, and suggested remediation steps. Uses computer vision (CV) image diffing on aerial/satellite images to spot physical site changes like excavation or clearing.
Predictive Risk Modeling: Employs machine learning to forecast future infractions before they happen (e.g., predicting a soil moisture breach within 72 hours or upcoming erosion risks).
AWS Tools we used: Uses Amazon SageMaker for AI-powered site visualization and Shap-E-based 3D generation. Amazon SQS queues terrain-generation requests, AWS Lambda handles terrain-processing tasks, and Amazon S3 stores elevation grids, generated models, and checkpoints.
Unified Dashboard & Reporting: Allows filtering by risk level (Low/Medium/High) or status (Compliant/Flagged/Critical), while generating weekly audit reports, exportable violation logs, and compliance scorecards.
How We Built It Frontend UI & Data Visualization: Built with React and Tailwind CSS featuring a custom dashboard, real-time alert feeds, metric progress cards (Solar/UV, Humidity, Soil pH), and interactive parcel search filters.
Geospatial Mapping Engine: Interactive map canvas supporting GeoJSON/KML polygon rendering, layer toggles, coordinate tracking, and automated risk heatmaps.
NLP Extraction Pipeline: Natural Language Processing models designed to parse legal PDF lease agreements and construct structured JSON compliance schemas.
Predictive Machine Learning Engine: RandomForestClassifier models trained on datasets including WeatherBench2 to analyze weather patterns, historical infractions, and seasonal satellite trends for predictive risk scoring.
Virtual Sensor Architecture: Simulated live ingestion pipeline pulling weather, soil quality, and satellite data streams mapped to spatial parcel coordinates.
Challenges We Ran Into Parsing Complex Legal Jargon: Standardizing unstructured lease PDF text into deterministic compliance rules required fine-tuning NLP prompts and schema extraction logic.
Geospatial & Telemetry Synchronization: Aligning multi-variable time-series telemetry (soil quality, temperature, precipitation) with precise spatial parcel boundary polygons in real-time.
Predictive Model Accuracy: Feature engineering on weather dataset features to reliably predict 72-hour soil limit breaches without generating excessive false-positive alerts.
Accomplishments That We're Proud Of Successfully reducing a 6–10 hour/week manual compliance overhead per parcel down to real-time automated tracking.
Combining NLP contract parsing, live sensor ingestion, and satellite computer vision into a single unified platform.
Designing a user experience that bridges the gap between SCADA engineers, legal teams, and land agents.
What We Learned How critical spatial context is for legal compliance—attaching a lease rule to a physical set of map coordinates changes how field teams operate.
Practical techniques for training tabular classifier models (RandomForestClassifier) on environmental datasets like WeatherBench2 to generate predictive risk alerts.
What's Next for parcelsite Direct SCADA System Integration: Auto-triggering equipment adjustments or shutoffs directly via SCADA pipelines when critical environmental thresholds are breached.
Expanded Drone & High-Res Satellite Pipelines: Enhancing computer vision image diffing for continuous, automated physical change detection on massive land tracts.
Automated Regulatory Filing: One-click submission of formatted compliance reports directly to county and state environmental agencies.
Built With
- amazon-web-services
- bedrock
- kiro
- lambda
- python
- s3
- sagemaker
- sqs
