Honest provides an edge‑vision AI platform that delivers calibrated confidence scores, auditable decisions, and explicit operational limits for public‑infrastructure and transit operators. Its config‑driven detector engines run on a single 8 GB Jetson‑class device, eliminating the need for data‑center GPUs, and support twenty‑six predefined use cases such as VMS safety intelligence.
Funding
Funding not disclosed
Founders
Product
Problem
Transit and public‑infrastructure operators need real‑time safety monitoring from existing CCTV feeds, but traditional cloud‑based video analytics introduce latency, privacy concerns, and high infrastructure costs, limiting timely intervention in high‑risk environments.
Solution
Honest delivers an edge‑vision AI platform that runs calibrated, auditable detector engines on a single 8 GB Jetson‑class device, eliminating the need for cloud GPUs. The config‑driven system supports 26 safety use cases—including VMS alarm routing, zone‑based rules, and multi‑camera tracking—while providing explicit confidence scores and operational limits. Inference completes in under 150 ms per frame, enabling alerts to reach operators before incidents unfold. Raw video remains on‑premises by default, with only structured events, approved clips, and telemetry leaving the site under policy control. The platform integrates with standard RTSP/ONVIF cameras and major VMS solutions (Genetec, Milestone, Avigilon) via REST APIs, webhooks, and SIEM connectors, allowing seamless incorporation into existing security workflows.
Target Audience
Primary customers are safety and operations teams at transit agencies, logistics hubs, warehouses, manufacturing plants, and other public‑sector facilities that rely on existing CCTV infrastructure for real‑time incident detection.
Features
- Config‑driven detector engines covering 26 safety scenarios on a single 8 GB Jetson or comparable industrial edge node
- Sub‑150 ms local inference with calibrated confidence scores and auditable decision logs
- On‑premises video processing; only structured events and policy‑approved clips are exported
- Multi‑camera parallel processing, cross‑camera object tracking, and zone‑based rule engine
- Native integration with RTSP/ONVIF streams and VMS platforms (Genetec, Milestone, Avigilon) via REST, webhooks, and SIEM connectors
- Policy‑based privacy firewall and end‑to‑end encryption for data leaving the facility
- Ruggedized hardware (Sentinel‑X) with IP67 enclosure, fan‑less design, and 64 GB LPDDR5 memory for continuous operation