Mantid provides AI software that runs directly on existing security cameras, turning them into continuous, on‑site supervisors that monitor, report, and act on incidents without sending video to the cloud. The edge‑first platform requires only a short setup call and can be customized with bespoke models trained on a client’s footage within weeks, enabling real‑time protection across thousands of cameras.
Funding
Funding not disclosed
Founders
Product
Problem
Many organizations rely on existing security camera installations for surveillance but lack real-time, actionable intelligence from the video feed. Without on-site processing, incidents often go unreported, and sending raw footage to the cloud raises privacy and bandwidth concerns.
Solution
Mantid provides AI video analytics software that runs directly on the cameras or an on‑premise edge box, turning each camera into a continuous supervisor. The edge‑first architecture processes video locally, generating real‑time alerts and automated actions while keeping raw footage on site. Mantid offers a library of pre‑trained safety and operations models (e.g., spill, smoke, PPE compliance, forklift safety) that can be deployed in days, and it can train custom models on customer footage within four weeks to address specific detection needs. Alerts, metadata, and selected clips are securely transmitted to a cloud service for review, reporting, and integration with existing security workflows.
Target Audience
Primary customers are safety, security, and operations teams at large enterprises such as airports, retail chains, manufacturing plants, and distribution centers that already have extensive CCTV deployments.
Features
- Edge‑first deployment runs inference on‑site, eliminating the need for cloud video uploads and reducing bandwidth usage
- Compatibility with 500+ camera models (Hikvision, Dahua, Axis, Bosch, Hanwha, any ONVIF/RTSP device) without hardware replacement
- Library of six out‑of‑the‑box safety models plus operational telemetry (spill, smoke, PPE, restricted areas, forklift safety, blocked exits, dock cycle, bay utilization)
- Custom model training on customer footage delivered in approximately four weeks
- Real‑time alerts and automated actions with metadata export; optional clip export for incident review
- Security and compliance features: AES‑256 at rest, TLS 1.3 in transit, SOC 2 controls, role‑based access and audit logging
- Simple pilot setup: 15‑minute call, 30‑day pilot on a single site, rapid deployment within days for library models