Processing, analysing, and predicting, a camera security system does everything in real time. How? M Saravanan from Quantum Sharqs explains to EFY’s Nidhi Agarwal how they leverage AI-ML and edge computing to do so.
Q. Can you give us an overview of your company and its core offerings?
A. We initially started by focusing on the healthcare sector, developing medical software, particularly clinical management solutions for physiotherapy clinics. In these setups, patient progress is typically tracked using daily pain scale assessments, and we introduced innovations in how this data is recorded, managed, and analysed through our platform.
Building on this, we developed Quantum VizionX, an image and video analytics-based device designed to work with existing surveillance systems, and today we offer six variants of this solution. These include a CCTV (closed-circuit television) health monitoring system that detects tampering, power loss, or visual obstruction and sends alerts; an intelligent facial recognition system. Our system converts existing CCTV cameras into recognition systems without additional hardware; smart human detection for tracking footfall and basic demographics; a multi-phase capture system for monitoring repeated or restricted area access; smart mask detection for compliance; and an automated headcount system that provides real-time insights such as occupancy, age group, and gender using camera feeds.
Q. Can customers choose individual features, or do they have to opt for all six variants together?
A. Customers can choose features based on specific requirements. While it is a single-core device, functionalities can be enabled as needed. If a client requires all six variants, we provide hardware with higher specifications, such as scaling the RAM (random access memory) from 4GB (Gigabytes) to 16GB to support simultaneous processing. This solution is already active within a leading Indian jewellery group, providing a real-time dashboard for camera status and detailed analytics, including gender and age distribution over specific time slots.
Q. Where is your hardware connected in the system?
A. Our device connects to the NVR (network video recorder) or DVR (digital video recorder). By integrating at this central point rather than individual cameras, the system receives and processes video data from all linked cameras immediately upon power-up.
Q. Is your solution compatible with all types of cameras and setups?
A. Yes, it is compatible with any camera brand or type connected through an NVR or DVR. The only requirement is an active internet connection to transmit processed analytics and notifications to the client.
Q. How does the system deliver analytics and alerts to users?
A. Once the AI (Artificial Intelligence) models are operational on our hardware, users access outputs via web pages, mobile applications, WhatsApp, or standard notifications. The system provides real-time updates on person counts and sends immediate alerts for suspicious activity.
Q. How is your system deployed across locations, and how does it scale across multiple cameras?
A. Deployment occurs wherever an NVR exists, such as traffic signals or police stations. In Tamil Nadu, stations in Tiruchengode and Pallipalayam use this setup to monitor networks without infrastructure changes. A single device can process four, eight, or up to 64 channels simultaneously. For larger installations with approximately 250 cameras, we deploy multiple devices in proportion to ensure scalability.
Q. How are cameras physically connected in such large deployments?
A. In large-scale setups, cameras installed across various locations are linked via fibre optics or other communication methods to a central monitoring centre. All feeds are aggregated at the NVR, where our device processes the data from a single point.
Q. Can your system work with smart or Wi-Fi cameras that use cloud storage?
A. For cameras using cloud or SD (secure digital) storage, our hardware does not require a physical connection; however, we would need authorised access to the encrypted video stream. As such, permissions are rarely granted; our solution is typically not used in these configurations.
Q. When pitching to organisations, does each location need its own device, and how is monitoring handled?
A. Yes, each location requires a device for centralised oversight. For a bank with 256 branches, a central authority can track the live status and exact location of every camera. Our health-monitoring variant is highly sought after in high-value sectors, such as jewellery, for its real-time alerts to tampering or power failures.
Q. Does your system provide AI and ML (machine learning) capabilities?
A. Yes. Unlike basic camera notifications limited to a single unit, our system scales to support over 64 cameras across multiple locations. It is vendor-agnostic, allowing seamless integration with mixed hardware brands such as Hikvision and Dahua.
Q. Is video processing handled on the device or in the cloud?
A. We offer both. Edge processing on the device is generally more cost-effective. Moving processing to the cloud requires additional infrastructure, such as a NAS (network attached storage) server, which can increase costs by approximately ₹100,000.
Q. Do your devices support over-the-air updates?
A. Yes, all devices support remote updates for software improvements and bug fixes, ensuring the system stays current without manual intervention.
Q. Can partners or third parties install the system?
A. Yes, the system is designed for easy integration by channel partners into existing infrastructures, though we provide support whenever necessary.
Q. Does the system enable district-level oversight of camera status?
A. Absolutely. By installing at key points like police stations, the system monitors all linked cameras in the network. If any unit is tampered with, the system automatically notifies the designated officer for effective regional oversight.
Q. What backend infrastructure supports device management?
A. We provide ongoing maintenance and conduct system checks every three months. The backend is designed to ensure the device remains consistently operational while handling real-time alerts.
Q. Who are your primary target users?
A. We target large organisations with 250 or more cameras, including enterprises, industrial facilities, and government networks. While we serve sectors like jewellery, small setups with few cameras are not our focus, as they often rely on basic built-in analytics.
Q. What feedback have you received from early customers?
A. Feedback has been positive regarding reliability. One challenge involves IP (Internet Protocol) cameras, where maintenance-related IP address changes can temporarily affect analytics. Otherwise, clients value the real-time monitoring performance.
Q. What other software products do you offer?
A. We offer three main products. Quantum VizionX is a hardware-software hybrid for computer vision analytics. Quantum VizionX is a comprehensive HRMS (Human Resource Management System) solution that combines both hardware and software components. It offers employee attendance management, facial recognition, access control, leave management, payroll support, and workforce analytics, enabling organisations to streamline and automate their HR operations. Quantum Healthcare (formerly Quantum Physio) is a clinical management ERP (Enterprise Resource Planning) system for physiotherapy and orthopaedic practices. It helps clinics manage their complete operations, including patient records, appointments, treatment plans, and progress tracking.
Q. What inspired the company, and is there a story behind the name?
A. The company was born from my technical background in computer vision during my PhD. We started in 2023 with medical software to generate initial revenue, which supported our transition into advanced hardware solutions like Quantum VizionX. We currently have a team of 15 employees.
Q. Do you manufacture all components in-house?
A. Not entirely. We source components such as GSM (Global System for Mobile Communications) modules and Raspberry Pi from external vendors. Our core innovation lies in the assembly and integration of our proprietary AI-driven software.
Q. Can you share details regarding your hardware development?
A. We currently use a single-board architecture with encrypted software. However, development boards include unnecessary components that keep costs high, around ₹13,000 to ₹14,000. We are moving toward a custom printed circuit board (PCB) design to optimise costs. Assembly and testing are conducted in-house at our Vellore office.
Q. Where is the product manufactured?
A. Components like Raspberry Pi are purchased from an authorised vendor in India, and the assembly is done in-house at our office in Vellore, Tamil Nadu. We assemble the hardware, design the casing for Quantum VizionX, integrate the AI chip or cloud-based software, and test the system before it is sold.
Q. How are AI and ML applied in Quantum VizionX?
A. The hardware runs a Linux-based OS (operating system) and uses frameworks like TensorFlow and OpenCV. We utilise CNN (convolutional neural network) models like YOLOv8 and SSD (single-shot detector) for real-time object and anomaly detection. Data is converted into structured JSON (JavaScript Object Notation) for alerts via the cloud or WhatsApp.
Q. Why develop a new board if Raspberry Pi works?
A. Our goal is incremental research to reduce costs and improve commercial viability. Choosing Raspberry Pi originally provided a balance of performance and price compared to expensive options like Jetson Nano or limited ones like ESP32 (Espressif System 32). Our custom PCB, expected by May 2026, could reduce component costs from ₹11,000 to approximately ₹4000.
Q. If Raspberry Pi and other development boards like Jetson Nano exist, why did you choose Raspberry Pi for your system?
A. We evaluated all options like ESP32, Raspberry Pi, and Jetson Nano, but each had limitations. ESP32 could not support RTSP (Real Time Streaming Protocol) camera streams, which are essential for our application. Jetson Nano, while capable, is too expensive for commercial deployment, costing around ₹40,000. Raspberry Pi offered the perfect balance, with a 4GB (Gigabytes) model at ₹13,000 and 2GB at ₹7000. It handles all required operations, supports RTSP streams, and keeps costs reasonable/
Q. Are you working on the next version of the device, and how will it differ from the current one?
A. As I said, we are enhancing the system with a custom PCB design expected by May 2026. By moving from third-party development boards to in-house manufacturing, we aim to reduce component costs from ₹11,000 to approximately ₹4000-₹5000; a reduction of over 50 per cent. While internal architecture will change, user functionality remains consistent. Additionally, we are proceeding with trademark registration to protect the core innovation of the device.
Q. What challenges do you face in scaling?
A. The primary hurdle is market penetration. Because our solution is unique, we must educate clients on the problem it solves. Growth is driven more by direct engagement than social media. We recently converted to a private limited entity, Quantum Shock Digital System Private Limited, which granted us Startup India certification and access to potential funding of up to ₹1 billion.
Q. How do you optimise bandwidth usage while ensuring security?
A. Access to video streams in India is strictly regulated; therefore, we access feeds exclusively for analytics rather than continuous monitoring. Integration with an NVR is secured via credentials and strict NDAs (Non-Disclosure Agreements). This configuration ensures real-time analysis, minimises bandwidth, and maintains full regulatory compliance.
Q. How do you test and verify your devices before deployment?
A. We perform six to nine months of rigorous in-house lab testing, simulating real-world scenarios such as lens obstruction, wire cutting, and internet disconnection. Every possible anomaly is tested to ensure reliable performance and accurate responses to tampering before commercial release.
Q. How accurate is the system, and have there been instances of incorrect results?
A. The core analytics and detection are nearly perfect. Any minor delays, typically 15 to 60 seconds. stem from network connectivity or communication protocols rather than inaccurate detection. The system consistently delivers accurate alerts within one minute of an incident.
Q. Where does the electronics aspect fit into your solution?
A. The electronics reside in the hardware layer, specifically the board and its components. While AI (Artificial Intelligence) and ML (Machine Learning) handle processing, we are also developing a cloud-based model. This will provide greater scalability and flexibility by shifting the compute load from physical devices to remote processing.
Q. What are the current challenges you are facing to grow superfast as a startup?
A. Our primary challenge is scaling through effective market penetration and positioning. While the technology is market-ready, growth is hampered by product perception and the difficulty of reaching the right customers. Since social media and general outreach have proven insufficient, we are pivoting toward direct client engagement to demonstrate operational value. Furthermore, limited marketing support impacts adoption; for instance, a technically sound proposal to monitor 4000 cameras for the Odisha government has been delayed. The critical gap remains effectively positioning and pitching the solution to enable large-scale deployment.
Q. How would you describe the competitive landscape in your domain?
A. Interestingly, the lack of direct competition itself is a challenge. Since there are very few, if any, comparable solutions in the market, we not only have to pitch the product but also educate potential clients about the problem it solves and how it differs from existing systems. For instance, in discussions with the Odisha government, we need to clearly communicate the value proposition; while traditional facial recognition systems focus on identifying individuals, our solution is designed to monitor large-scale networks, such as an entire city, and provide real-time alerts if any camera is interrupted or tampered with.
In terms of real-world deployment, one of our users, R. Kandan, has been using the system for the past nine months, demonstrating its reliability and scalability from a single installation to broader surveillance applications, as envisioned for projects like Odisha.
Q. Are you currently receiving any government or external funding?
A. Until recently, we were not receiving any funding, as the company was registered as a proprietorship under Quantum Shack Innovate Solution for the past two years, which led to repeated rejections for funding, including Startup India support, and also made it difficult to secure bank loans or state government assistance. However, last month in 2025, we converted the company into a private limited entity, now named Quantum Shock Digital System Private Limited, and have since received the necessary approvals, including the Startup India certification, which I already mentioned.
Q. Where are your current investment and partnership focuses?
A. We are focusing on direct marketing to key decision-makers, including CEOs of HDFC (Housing Development Finance Corporation) and CU (City Union) Bank, and DGPs (Directors General of Police) across multiple states. We are also actively seeking channel partners in camera manufacturing and sales to expand our distribution and support network. We also welcome collaborations from adjacent sectors, such as our upcoming partnership with an emerging group in Tamil Nadu.
Q. What are your future growth plans?
A. Our focus remains on consultative marketing. By educating sectors on the unique capabilities of our analytics, we aim to drive large-scale adoption across various industries. To date, we have sold 47 units of Quantum VizionX, generating ₹1.9 million in revenue.
