Architecture
Six layers from sensor input to mission record.
EdgeLance is not a dashboard on top of third-party APIs. It is a vertically integrated platform where sensing, AI, mesh, operations, device management, and review are built as one system.
6
Execution layers
7+
Mesh link types
4
Classification levels
5
Dashboard views
6
Sensor modalities
5
Compute routing policies
4
Node agent platforms
28+
TPMS protocols
Technical Stack
Built on proven, portable foundations.
Sense
Ingest and normalize field inputs from cameras, RF receivers, NFC readers, acoustic sensors, and partner feeds into structured mission events.
Camera Service
RTSP ingestion, zone detection, MQTT publishing, MJPEG streaming. Pluggable object detection pipeline tuned for low-latency edge execution.
Face Service
Enrolled personnel matching with NFC badge corroboration. Returns resolved identity with role, clearance, blood type, casualty status.
RF/TPMS Service
RTL-SDR monitoring across 28+ TPMS protocols. Passive vehicle fingerprinting at range from tire pressure sensor broadcasts.
Acoustic Service
FFmpeg audio extraction from RTSP. dB monitoring, loud event detection, voice activity, footfall analysis, and local speech-to-text.
LPR Service
Plate extraction via detection pipeline and OCR. Watchlist cross-reference with confidence scoring and hot-list alerts.
Understand
Run local AI inference to classify, assess, and contextualize sensor data into actionable intelligence with full evidence coupling.
Threat Analyzer
On-device language model inference on Apple Silicon and NVIDIA GPUs. Event-driven: triggers on scene changes only. Returns classification, confidence, recommended actions, intercept priority. Supports models like Gemma, Llama, Mistral, or customer-approved alternatives.
Video Orchestrator
Vision-language pipeline routing. Detection triggers deeper analysis selectively. Cached results minimize redundant compute across the model stack.
Segmentation Engine
On-demand mask generation for scene understanding, perimeter breach analysis, and precision object isolation. Supports SAM-class models and alternatives.
Compute Policy Engine
Routes inference by data sensitivity, device capability, and mission policy. Supports local-first, base GPU, cloud, classification-gated, and DDIL-only routing modes.
Model Catalog
Version tracking, approval status, export restrictions, approved use classifications. Full provenance from training to deployment.
Route
Move the right data to the right node through degraded or disconnected networks using mission-aware mesh routing.
Mesh Router
Opportunistic multi-hop across WiFi, LoRa, Starlink, LTE, satellite phone, BLE, SBD. Route selection co-optimizes bandwidth, latency, and power.
Link Monitor
Per-link latency, packet loss, jitter, bandwidth tracking. Link grading for route selection. Flapping detection and penalization.
Store-and-Forward
Event queuing during link loss. Priority-based replay on reconnect. Snowball effect: data quality upgrades at each better-link hop.
Bandwidth Scheduler
Camera claim system allocating bandwidth by tactical priority. High-priority streams get full resolution. Low-priority streams drop to thumbnails.
Remote Relay
Cloudflare Workers edge relay. WebSocket connections for disconnected teams. Short-term event queue supports tactical link disruption and replay.
Act
Present mission-relevant information and actions to each role through purpose-built operator views.
PLAN View
Mission readiness assessment, sensor registration, route/role setup, model loadout selection, launch gate validation.
LIVE View
Real-time entity map, multi-camera wall with AI overlays, threat picture, audio transcript, contact log, mobile node tasking.
REVIEW View
Chronological event timeline, 10-second evidence clips (H.264+AAC), searchable contact history, local AAR generation, export/lock.
Mission Agent
Model-agnostic tactical agent converting operator directives into auditable action plans. Auto-executes ISR/support steps, gates decisions with ROE checks.
Voice Command
Speech-to-text intent parsing and action execution. Hands-free tactical commands with confirmation gates.
Control
Manage devices, models, software, and security posture as part of the mission workflow.
Classification-Aware MDM
Configurable classification tiers with distinct profiles, data boundaries, auto-wipe timers. Wraps open-source MDM infrastructure with EdgeLance enforcement.
STIG Manager
DISA STIG checks mapped to continuous device query policies. Real-time evaluation, not audit-based. Passcode, auto-lock, USB, VPN, AirDrop/iCloud controls per level.
Tactical Features
Stealth Mode (radio and network suppression controls), NVG Mode (green monochrome, night-vision-aware), Duress PIN (silent wipe, covert alert, hardware-backed secrets).
Fleet Management
Zarf packages for airgapped delivery. Software Courier via managed iOS. Cross-platform node agents for macOS, Windows, Linux, Android.
Product Spine
Central governance registry. Enrollment sessions, model approvals, readiness reports, rollback points, device receipts, fleet rollouts. Full audit chain.
Review
Preserve the mission record for command, legal, training, partners, and next-mission learning.
AAR Service
Auto-generated narrative reports grouping events by entity/incident. Local text generation with a structured format (who/what/when/where/why).
Evidence Store
SQLite event database for offline replay and forensics. Queryable by time window, entity, incident, threat level.
Replay System
Full mission recording to disk for offline playback and training. Enables rehearsal of complex scenarios.
Audit Logging
Every action logged: sensor events, threat assessments, response authorizations, MDM changes. Operator, timestamp, hash, signature. JSONL + MQTT.
TAK Publishing
Entities converted to Cursor-on-Target XML. Published to TAK Server or peer-to-peer mesh via UDP multicast or TCP. Designed for ATAK/iTAK/WinTAK interoperability where approved.
Security Model
Defense-grade posture on consumer hardware.
Security is not a feature layer. It is enforced at every level: device, network, data, inference, and audit.
Data classification enforcement
Compute policy prevents classified data egress. Inference routing gated by classification level, device type, and mission context.
Hardware-backed secrets
Duress PIN, device certificates, and encryption keys use hardware-backed storage where supported. Designed to reduce exposure during device compromise.
Continuous STIG compliance
Device query policies evaluate DISA STIG controls in real time. Drift detection triggers alert, restrict, or wipe based on severity.
Audit segregation
Logs from different classification levels stored and transmitted through separate channels. No cross-level contamination.
MQTT authentication
Broker-level credential enforcement across all sensor services and event channels.
Mission evidence integrity
SQLite event database with hash chains. Tamper detection on replayed evidence.
Deployment
One platform. Multiple deployment models.
Tactical Edge
MacBooks, iPhones, iPads, NVIDIA Jetson, IP cameras. Fully airgapped. Software Courier for updates.
Base/FOB
Kubernetes cluster with Helm charts. GPU inference servers. High-bandwidth mesh backbone.
Hybrid
Edge nodes with cloud compute offload when policy and bandwidth allow. Seamless handoff.
Training/Demo
Simulation engine generates realistic mission data. Full capability without live sensors.
Walk through the architecture, not a slide deck.
Request a technical deep-dive with our engineering team.
Request Demo