PixelSight

**Product and the concrete problem it solves.** PixelSight is addressing the warning gap created by small, low, quiet, and visually ambiguous threats. A radar or RF system may tell an operator that something is present, but many security decisions require knowing whether the object is a drone, bird, vehicle, person, or another anomaly, and doing so early enough to change the outcome. The company’s public product architecture has two layers. **EdgeSight** is the field-deployed sensing system: a sensor unit and embedded detection software that continuously observes the environment, detects objects in real time, raises alerts, and classifies targets. **OmniSight** is the intelligence layer: it collects and fuses visual data from deployed EdgeSight systems, produces a unified operational picture, and supports a detection-to-decision workflow while keeping the human operator as final authority. The immediate use case is counter-small-UAS warning, but the platform is framed for multi-domain awareness across land, low-altitude airspace, and maritime environments. The operational problem is not simply finding pixels; it is reducing the time and false-alert burden between a physical event and a confident response.

**Core technology and how it works.** PixelSight describes a passive system with zero RF emissions, an important design choice for operators that need to observe without advertising the location of the sensor. Its public product page says the sensing layer extracts the physical signature of extremely small and low-signature threats, performs early filtering and alerting, and can distinguish categories including drones, birds, vehicles, and humans. OmniSight then performs multi-sensor data fusion, deeper validation, location and route estimation, and false-alarm reduction across deployed systems. The company does not publish the camera make, optical baseline, model architecture, training-data scale, detection probability, or environmental test protocol, so this record should not infer a proprietary sensor modality or a specific neural-network method. Its technical hiring material is more revealing about the engineering boundary: PixelSight recruits for Linux and edge-system work, Dockerized deployments, Python and Bash automation, OpenCV, object detection, PyTorch, TensorRT, CUDA, ONNX Runtime, NVIDIA Jetson and ARM64 devices, camera integration through industrial SDKs, ONVIF and RTSP, and 3D geometry such as calibration and coordinate transforms. That combination indicates a hardware-software system designed to run inference close to the sensor, where latency, power, connectivity, configuration, and field reliability matter as much as model accuracy.

**Market, customers, and go-to-market.** PixelSight is selling into a security market in which buyers increasingly want layered detection rather than a single sensor type. The Israel Innovation Authority identifies defense and security organizations, border and airspace-security operators, critical-infrastructure operators, maritime-security teams, and public-safety organizations as target customers. The company’s own positioning broadens the commercial route to civil safety and other environments where early warning and high-confidence classification have economic value. A plausible go-to-market sequence is to start with defense and government-backed operational pilots, prove detection performance in difficult conditions, and then use the same passive optical layer in airports, ports, utilities, industrial sites, and protected perimeters. This is likely a systems sale, not a self-serve software motion: installation geometry, sensor placement, communications, edge-compute configuration, integration with command-and-control or video-management systems, and ongoing field support will influence price and gross margin. Public sources do not disclose recurring revenue, named commercial customers, contract values, channel partners, or the split between product revenue and deployment services. The public hiring page does, however, say the systems are already deployed in critical environments and that the company is hiring across field systems, applied research, procurement, and end-to-end integration, which is consistent with an organization moving from prototype work toward repeatable deployment.

**Traction, funding, and third-party validation.** The strongest disclosed validation is government-linked rather than a large venture round. The Israel Innovation Authority’s public company page lists PixelSight as established in 2024, with 50 employees, a revenue-growth stage, and support through an applied-research program in collaboration with the Directorate of Defense Research and Development (MAFAT), alongside technological-incubator support. The same page says the technology can detect attack drones at distances above three kilometers and names Omer Cahana as CEO and co-founder and Eyal Krupka as CTO and co-founder. Startupim reports undisclosed Seed and later financing entries, including participation by Cavaret and Earth & Beyond Ventures, but the amount and precise round history are not independently confirmed here. In July 2026, iHLS and other outlets reported that the Israeli Ministry of Defense and IDF acquired PixelSight drone-detection systems for field evaluation; the reported purpose was to test suitability for operational integration, not to claim a completed scaled procurement or permanent deployment. The company’s own hiring page adds a useful operating signal by stating that its vision-intelligence platform is deployed in critical environments and that its systems are intended to work in production and field conditions. These signals are meaningful, but diligence still needs measured detection and false-alarm results, signed procurement documentation, customer references, revenue recognition, and confirmation of the current employee count.

**Founders and team background.** PixelSight was founded by Omer Cahana and Eyal Krupka. The Innovation Authority identifies Cahana as CEO and co-founder and Krupka as CTO and co-founder; company-registry reporting identifies the legal entity as Pixel Sight Ltd., incorporated in January 2024 in Herzliya. Public professional material is limited, but independent company information describes Krupka as having previously worked at Microsoft, while Cahana’s public profile connects him with Google Research and earlier Bosch Research experience. The combination is relevant to the product: computer-vision research and production engineering are necessary, but the company also has to build optics, embedded systems, communications, power, calibration, field-test processes, procurement, and deployment operations. PixelSight’s open roles explicitly cover Linux edge systems, senior computer vision and real-time systems, applied research, systems engineering, procurement, and production readiness. That hiring pattern suggests a multidisciplinary team rather than a model-only startup. The caveat is that biographies, exact headcount, organizational depth, and prior defense-program experience are not fully disclosed in public sources. Team score should therefore recognize unusually strong founder-to-problem fit and evidence of a real systems organization, while reserving credit for military-procurement expertise, classified-program access, and sustained international delivery until those facts are documented.

**Competitive dynamics.** PixelSight competes in a layered counter-UAS and perimeter-sensing stack, not in an empty category. Local alternatives include ThirdEye Systems’ EO/IR recognition products, Axon Vision’s passive computer-vision and counter-UAS systems, and Magos Systems’ radar-plus-AI perimeter sensing. Global alternatives include Dedrone’s airspace-security platform, Robin Radar Systems’ bird-and-drone radar, Teledyne FLIR Defense’s EO/IR payloads, and defense-prime solutions from Rafael and Elbit Systems. There is also a non-technology substitute: combine radar, RF detection, acoustic sensors, human watchstanders, and camera verification through an integrator. PixelSight’s potential edge is the end-to-end pairing of a passive sensing layer with a separate intelligence layer that can fuse multiple deployments, reduce false alerts, and support a common operational picture. It also claims a low-cost, defense-grade positioning and broad land-air-sea coverage. Those claims matter only if independently demonstrated. The decisive comparison will be probability of detection for small targets, time to alert, classification precision, performance in darkness, haze, rain, clutter, camouflage, and changing backgrounds, operation against RF-silent or fiber-connected drones, resilience to adversarial deception, power consumption, cyber-hardening, and total installed cost. A three-kilometer claim without a target profile, weather condition, field-of-view, and false-alarm denominator is a useful signal, not a benchmark.

**Defense, security, and resilience dual-use relevance.** PixelSight merits dual_use=true because the core technology is already framed for defense and public safety, while the sensing and analytics architecture has credible civilian applications. Passive optical detection can complement radar and RF layers around military bases, border areas, ports, airports, energy installations, water facilities, and other critical sites. Its relevance is especially clear for low-altitude, low-signature, or RF-silent drones that may evade a single modality. OmniSight’s multi-sensor fusion and human-authorized detection-to-decision workflow could also support emergency response, search and rescue, maritime safety, wildfire or disaster monitoring, and protection of industrial sites where a false alarm is costly but a late warning is worse. The strategic value is therefore in resilient situational awareness and sensor-layer diversity, not in claiming that PixelSight itself provides interception or autonomous weapons. Public reporting of IDF and Ministry of Defense acquisition for field evaluation converts the defense thesis from pure adjacency into early government validation, but it does not establish combat deployment, a production contract, a classified program, or an operational kill chain. Important diligence questions include whether the system continues to perform when GPS, communications, or cloud services are degraded; whether it is cyber-secured and maintainable in isolated networks; whether the company holds the necessary export approvals; and whether the same models generalize across Israeli, allied, and commercial environments.

**Growth stage, trajectory, and key diligence risks.** PixelSight is classified as **mid**: it is a young company founded in 2024, but public records indicate a 50-person organization, revenue-growth status, MAFAT-linked research support, a deployed platform, and acquisition of systems for IDF field evaluation. That is materially beyond an idea-stage company, while still short of mature status because funding amounts, recurring revenue, customer concentration, production scale, retention, international certifications, and completed operational procurement are not public. The trajectory is attractive if PixelSight can become the optical classification layer that plugs into heterogeneous counter-UAS and critical-infrastructure architectures rather than a single-project sensor vendor. The main risks are: (1) **technical proof risk**, because public performance claims lack test methodology and comparative false-alarm data; (2) **category risk**, because radar, RF, thermal, acoustic, and integrated-defense competitors can bundle overlapping capabilities; (3) **field reliability risk**, because optics, edge compute, calibration, connectivity, and power fail in ways that a laboratory model does not reveal; (4) **procurement risk**, because defense and public-safety sales are slow, lumpy, regulated, and reference-driven; (5) **capital and manufacturing risk**, because scaling ruggedized hardware and field support is more expensive than shipping software; (6) **security and export risk**, because defense data, sensor locations, and algorithms require strict controls; and (7) **disclosure risk**, because public funding, revenue, headcount, contract, and patent information is incomplete. The next milestones to watch are a named production customer, independently measured field benchmarks, repeat orders outside Israel, documented integration with existing command systems, and evidence that OmniSight’s network effect improves performance across deployments.