ClearLease - AI Lease Abstraction Tool

Engineered a standalone, full-stack web application designed to automate commercial lease abstraction at institutional depth. Built with a FastAPI backend and a Next.js/Tailwind frontend, the platform processes complex legal leases, extracting over 260 custom data fields with unmatched speed, auditability, and data security. Unlike legacy tools that rely on manual abstraction or basic OCR, the system utilizes Azure Document Intelligence to transcribe handwritten notes, amendments, and margin annotations, and dynamically routes clean digital PDFs to bypass unnecessary processing. Users can build a custom extraction template right inside the app by choosing from 260+ available fields, or upload their firm's existing Excel, Word, or PDF lease-abstraction template and let the system detect the key terms and structure on its own to build a matching template from it. From there, every future extraction exports straight back into that same format, so the output looks exactly like what the firm already expects, no manual reformatting required. Every extracted data point also includes transparent confidence scoring and an interactive source citation, so users can jump directly to the exact highlighted clause on the original document, and anything short of full confidence gets flagged for a quick human check instead of being taken on faith.

• • Advanced OCR & Handwriting Recognition: Leverages Azure Document Intelligence to read handwritten amendments, broker annotations, and scanned image files, eliminating the blind spots of legacy OCR tools.
• • Interactive Synced Source Citation: Links every extracted field directly back to its source page and paragraph, featuring an interactive viewer that highlights the precise clause on the original document for total auditability.
• • Graduated Confidence Scoring: Automatically calculates and flags confidence levels across a 5-tier spectrum (from Verified to Needs Review), transforming manual due-diligence review into an efficient, targeted audit.
• • Smart OCR Pipeline Routing: Automatically detects whether a document contains an embedded text layer, bypassing OCR for clean digital PDFs to optimize processing speed and reduce operational latency.
• • Custom & Native Template Ingestion: Builds a reusable extraction template from 260+ available fields inside the app, or ingests a firm's existing Excel, Word, or fillable PDF template and detects its key terms automatically, so every export mirrors that exact native layout going forward.
• • Batched, Concurrent LLM Extraction: Groups independent fields into concurrent batch requests using Claude 3.5 Sonnet, accelerating full 260+ field extraction runs in a fraction of sequential call times.
• • Zero-Data-Retention Privacy Architecture: Engineered with strict privacy guardrails, no user data, uploaded leases, or extracted terms are saved, sold, or used to train external models.