4  Product Roadmap

Kav AI Platform Release Horizons & Product Capabilities

This document presents the product-level roadmap and release horizons for the Kav AI Platform (KAP). Aligned with our living Product Requirements Document (PRD), this roadmap maps the product vision onto chronological milestone horizons from May 2025 through the end of 2026 and beyond.

5 Product Vision and Pillars

Our roadmap is structured around four primary product pillars that together deliver Active Physical Intelligence™:

  • 3D Spatial Reality (The Digital Core): Merging raw reality-capture and engineering drawings into a single, high-fidelity, interactive 3D representation of the physical facility.
  • Multi-Modal AI & Intelligence: Applying agentic models to reason across visual (RGB), infrared (thermal), optical gas imaging (OGI), and historical process data to identify anomalies.
  • Connectivity & SCADA Integrations: Building read-only OT bridges to SCADA, historians, and P&ID databases to cross-reference real-time signals with visual findings.
  • Enterprise Security & Compliance: Providing multi-tenant role isolation, SOC 2 Type II assurance, audit-ready environmental records, and self-hosted air-gapped deployment profiles.

6 Completed Release Horizons

6.1 Milestone M0: Platform Foundation (Jul 2025)

  • Product Theme: Bootstrap and Core Visualization
  • Status: Complete / Delivered

Key Capabilities Delivered:

  • CesiumJS 3D Viewer: Basic high-fidelity web-based rendering of 3D spatial capturing.
  • Operator Dashboard: Entry point for inspectors and maintenance managers.
  • Multi-Tenant Auth: Secure user access control and multi-tenant database isolation.
  • Visual Defect Gallery: Basic image review panel with metadata and categorization.

Validation: Verified and field-tested using a first visual inspection campaign consisting solely of RGB imagery.

6.2 Milestone M1: AI Foundation (Dec 2025)

  • Product Theme: On-Demand defection and Natural Language Interface
  • Status: Complete / Delivered

Key Capabilities Delivered:

  • Multimodal AI Pipeline: AI framework combining visual analysis with structured data reasoning.
  • Natural Language Chat: Conversational interface enabling operators to ask simple questions of their data.
  • Automated Agentic Coordination: Multi-agent task dispatch and scheduler to orchestrate background analysis.
  • On-Demand Defect Detection: Proof-of-concept visual perception pipeline called by the conversational agent.

Validation: Tested against a second inspection campaign incorporating thermal imagery and structured environmental datasets (humidity, temp, gas).

6.3 Milestone M2: App MVP (Mar 2026)

  • Product Theme: Single-Screen Operator Experience
  • Status: Complete / Delivered (Released under v0.2.0 on April 25, 2026)

Key Capabilities Delivered:

  • Unified Interface: Side-by-side layout locking the photorealistic 3D viewer directly to the AI Chat Copilot.
  • PhotoMap3D Restructuring: Adoption of the modular, high-stability React feature-module folder architecture.
  • Inspector Notes Workflow: End-to-end notes triage browser with compact status controls and angled fly-to view.
  • Vitest Three-Tier Test Suite: Resolution of JS memory heap bottlenecks and validation of 95% of test-suite goals.

Validation: Validated against a dataset scanning 1,200 assets, flagging 37 anomalies (with 8 critical defects representing $750K–$1.5M in avoided damage).

7 Active Release Horizon

7.1 Milestone M3: AI Q2 Delivery (Jun 2026)

  • Product Theme: Conversational Access and Agent Reliability
  • Status: In Progress (Active shipment; target release is v0.3.0 on May 27, 2026)

Key Capabilities Shipping:

  • Contextual Data Chat (Phase 0): Workspace-scoped, multi-turn AI interactions (C0–C5 modes) with context-aware chips and state persistence.
  • Modular Workspace Switcher: Dual-route shell (Data Explorer & Integrity workflows) supporting custom role-gated personas.
  • Failure Recovery: Resilient chat sessions — message queueing and retry across network interruptions and agent failures.
  • Agent Evaluation Tests: Automated end-to-end agent evaluation suite (correctness and regression) gating each release.
  • 3D CAD Model Ingestion: Direct support for Navisworks-aligned 3D Tilesets and CAD-overlaid views.
  • Splat & Mesh 3D Alignment: Precision transform controls (lat/lon/height numeric sliders) to orient 3D captures manually.

8 Future Release Horizons

8.1 Milestone M4: Persistent Sensing (Q3 2026)

  • Product Theme: Multi-Sensor Detection and Engineering Context
  • Status: Planned / Upcoming

Q3 2026 deepens detection and grounds it in open-standard engineering context.

  • Multi-Sensor Automated Detection: Continuous AI anomaly categorization running on the backend across OGI, gas, and thermal feeds.
  • CAD (IFC) Ingestion & P&ID Tag-Linkage: Open-standard engineering context — CAD geometry via IFC4 with logical P&IDs via DEXPI and deterministic dual-tagging.
  • Geo-Tagged Spatial Anchoring: Ingesting GIS datasets, anchoring inspection campaigns, and displaying asset photos in 3D space.
  • Contextual Chat Phase 1: Compound queries traversing both visual databases and live data sources, with human-in-the-loop (HITL) safety gates.
  • Rescheduled from Q2 (M3 → M4): Sensor-native agentic analysis, actionable insights, chat with 3D map, interactive overlays, and automated reports — deferred as M3 narrowed to contextual chat, persona workspaces, failure recovery, and agent evaluation.

8.2 Milestone M5: Engineering Context & Enterprise (Q4 2026)

  • Product Theme: Engineering Context, Enterprise Compliance, and Deployment Flexibility
  • Status: Planned / Upcoming

Q4 2026 completes the platform and hardens it for enterprise deployment.

  • Read-Only SCADA Connectors: Telemetry and process historian ingestion via OPC UA (IEC 62541), mapping live signals to 3D assets.
  • SCADA UI Dashboard: Trend analysis and warning thresholds for process variables directly inside the Web App.
  • As-Built vs. As-Designed CAD Overlay: Overlaying actual design engineering files (IFC, RVT, DGN) over photorealistic models to calculate deviations.
  • P&ID SQL Connector: Connecting read-only to plant piping databases to sync engineering metadata.
  • SOC 2 Type II Certification: Formal audit completion meeting standard governance gates of global operators.
  • Self-Hosted Air-Gapped Deployment: Container-based offline deployment package with local LLM and model inference.
  • Compliance & Audit Workflows: Tracking overall inspection coverage metrics and generating records for regulatory submissions (e.g., EPA).

9 Long-Term Horizon (H1 2027 and Beyond)

Directional, beyond the committed 2026 milestones (M4–M5).

  • Autonomous Patrol & Navigation: Fixed-infrastructure robotics — navigation beacons, communication backbone, coverage orchestration, and fleet analytics — with full spatial navigation, for repeatable robot-patrol continuity decoupled from onboard SLAM.
  • Dose-Aware Operations: High-hazard environments package tracking radiation dosimetry and routing robot paths to avoid high-dose zones.
  • Regulated Deployment Profile: High-security on-premise profiles with customizable lifetime data retention policies.