.png&w=256&q=75)
Akira Dheena@akiradheena
🤝 Top Collaborators
.png&w=640&q=75)
🤓 Latest Submissions
NeoForge
NeoForge automates the end-to-end Intake → Understand → Decide → Review → Deliver lifecycle with surgical precision. It accepts real-world, multi-format inputs - PDFs (scanned forms with handwritten notes), CSVs (batch exports with missing fields/duplicates), and JSON/email submissions and normalizes them into structured, confidence-scored records. Using OCR and AI agents, it extracts key fields (claim ID, amount, injury type) while assigning per-field confidence (🟢 ≥0.9, 🟠 0.7–0.89, 🔴 <0.7) for risk-aware routing. Decisions cleanly separate deterministic rules (e.g., amount > $10,000) from AI reasoning (e.g., injury severity). High-value or low-confidence claims escalate to human review with full context (raw doc + data + rationale); missing-field cases trigger clarification loops — auto-emails, reply ingestion, reprocessing turning exceptions into recoverable workflows. Every action is fully auditable. NeoForge generates rich PDF audit reports (CLM789_Approved.pdf, CLM791_Rejected.pdf) featuring: • Input provenance (file, timestamp, size) • Extracted fields + confidence heatmaps 🟢🟠🔴 • Rules fired + AI rationale • Human review actions (with override reasoning) • Delivery status (Google Sheet row, email confirmation) • Full traceability: Job ID, workflow name, trigger source, GitHub audit path A lightweight Replit dashboard provides real-time status (✅ Approved, 🔄 In Review, ⚠️ Needs Info), QR-code access, and job history proving automation need not sacrifice visibility. Demonstrated via insurance claims (mock data: duplicate IDs, missing signatures, handwritten notes), NeoForge is modular and prompt-swappable instantly reusable for KYC, HR screening, or e-commerce returns. In an era where poor data quality costs millions, NeoForge delivers traceability, resilience, and trust transforming chaotic intake into confident, auditable outcomes. Built in 3 days. Designed to win.
19 Nov 2025
