Skip to main content

ITmatterss

From searching records to reading real-time intelligence: ValuePitch is reimagining the future of AI

Vertical Share Bar
From searching records to reading real-time intelligence: ValuePitch is reimagining the future of AI

Tech Talk with Venkat Ramana, CEO & Founder, ValuePitch

In a world increasingly shaped by digital footprints, fragmented databases and ubiquitous surveillance cameras, information is rarely the problem. Making sense of it, connecting the dots and turning it into actionable intelligence is the real challenge. This is the space ValuePitch has spent nearly two decades navigating. What began in 2006 has evolved into an AI-driven intelligence ecosystem built around a fundamental question: who is this person, and what can verified information tell us about them? From background verification and court-record searches to AML screening and digital identity checks, the company has built technology that helps businesses and institutions make more informed decisions.

Now, ValuePitch is taking that expertise beyond databases and into the world of video intelligence. Through NthEye, the company aims to transform conventional CCTV infrastructure from passive recording systems into searchable, AI-powered intelligence networks. Investigators can potentially trace individuals across cameras, identify possible matches and receive real-time alerts, reducing the hours spent manually reviewing footage. In a conversation with ITmatters, Venkat Ramana, the founder of ValuePitch, reflects on his journey and on a long-term bet on identity intelligence, built through years of institutional partnerships and largely bootstrapped innovation. With over 100 million searches to its credit, the company is now exploring a bigger question: can AI help make public safety more proactive without compromising individual privacy and accountability?

From ICWAI and finance to building Verify24x7, Belfort, Lendingcore and now NthEye and ValuePitch, you’ve multiple ventures. How did you come up with ValuePitch back then? How has it evolved over the years? What problem did you set out to solve in the beginning, and what kept you going through the years?

    I trained as a cost accountant (ICWAI) and began my career in finance and operations. ValuePitch started in 2006 as a sales and marketing services business. In that work, one problem kept coming up. Every lending, hiring and partnership decision came down to the same questions: is this person who they say they are, and is there a history I should know about? Nobody could answer them reliably.

    Court records were spread across thousands of courts, and most were on paper or on separate portals. By 2008 we had turned the company towards that problem. We built a data-mining platform to bring those records together and make them searchable.

    We grew in layers. First came background verification and court-record search for enterprises, through Verify24x7. Next came compliance tools for banks and financial institutions: AML and PMLA screening, related-party search and adverse-media checks. Belfort is our go-to-market product for US and B2B customers. Lendingcore is a lending infrastructure. Most recently, NthEye applied the same identity question to video.

    NthEye started with something personal. In November 2016, CCTV at a creche in Navi Mumbai showed a caretaker assaulting a 10-month-old girl, who was badly injured. The camera had recorded everything, but nobody saw the footage until after the child was hurt. My own son had just joined a creche. I was shocked, and I was scared.

    I put a team together to find out whether a camera could raise an alert while something was happening, not just record it for later. We worked on it for a year and failed. We tried again and failed again. Our fourth attempt began last year, and this year it finally worked.

    That is still what drives me. As a parent, I keep hearing stories about schools, buses and every other place our children go. A scared parent should be able to know when something is wrong while there is still time to act.

    The slow years were hard. We were digitising and matching court records before most buyers had a budget for it. We have been bootstrapped throughout, so we grew only as fast as customers paid us. That kept us disciplined. We built what people would pay for, and the customers who adopted it gave us the reason to keep going.

    ValuePitch works with key institutional stakeholders. What did it take to win the trust of police and government bodies for a private-sector product, especially court records?

    It took three things:

    First, we proved it on their cases. Police don’t buy after a presentation. They buy after a pilot where an officer sees the system work on a real case. Most of our relationships started as a proof of concept with one unit, run alongside the methods they already used, and grew from there.

    Second, we worked on their terms. The department decides who has access, every search is logged against an officer and a case reference, and senior officers can review that trail at any time.

    Third, we handled court records carefully. We work only with records the courts have already made public: judgments, orders and case listings. We present them for what they are, a record that a case exists, not a judgement on the person. A pending case is not a conviction, and our reports say so.

    After that, trust comes from consistency: the same team, support through every system change, and no overpromising.

    ValuePitch’s solutions span background verification, court record searches, AML screening and digital identity checks, and now AI video intelligence through NthEye. In simple terms, how do these technologies work and fit together? Why is identity the common thread?

    Every product we build answers one question: who is this person, and what do verified sources say about them?

    1. Background verification checks documents, employment and education.
    2. Court-record search checks for litigation across courts.
    3. AML screening checks sanctions lists, politically exposed persons and adverse news.
    4. Digital identity checks confirm that a person is real and is who they claim to be.
    5. NthEye asks the same question when all you have is a face in CCTV footage.

    They all work the same way. Each takes an identifier (a name, an ID or a face), matches it against trusted sources, and gives a person a ranked set of results to verify. The system narrows the search, and a human makes the decision.

    Identity is the common thread because almost every fraud, crime or bad hire starts with someone being someone they are not. Each product runs as a separate business for its own customers.

    With 100 million+ searches, what actually happens when an officer or enterprise runs a search? What data sources does it draw on, how does the system match a person or a face, and how has real-world usage made the engine more accurate?

    The user enters what they have. That might be a name with father’s name and address, an ID number, or a photograph.

    The system first normalises the input. Indian names come in many spellings, transliterations and initial formats, so “K. Srinivasa Rao” and “Kotha Srinivas Rao” have to be considered together. It then searches only the sources that user is authorised to see. For enterprises, those are public court records, regulatory and sanctions lists, and published news. For police, they are the records that department is permitted to use, accessed under its own authority.

    Name matching combines fuzzy matching with context such as father’s name, age and district to rank likely candidates. Face matching turns a face into a numerical template and compares it with templates in the authorised gallery. Both return ranked candidates with similarity scores, never a single “answer”.

    Real-world use is what improved the engine. Across more than 100 million searches, we learned Indian naming patterns, regional spellings and common aliases, and how to handle poor-quality, high-angle CCTV images. Each time an investigator confirms or rejects a candidate, we learn where to tune thresholds and matching rules.

    India has a huge amount of CCTV footage that is mostly only reviewed after something goes wrong. How does NthEye make that footage searchable, and how do face matching, suspect tracking and real-time alerts work in practice on existing cameras?

    Most CCTV’s in India are like a recorder in a corner. Footage is looked at only after an incident, by someone fast-forwarding through hours of video. We built NthEye so that a camera can alert people, not just record.

    NthEye works on top of existing cameras and recorders. As footage comes in, it detects and timestamps people, faces, vehicles and attributes such as clothing colour. That turns video into something you can search like a database: “show me this person across these cameras between 8 and 11 pm.”

    1. Facematching: an investigator uploads a photo of a suspect, and the system returns possible appearances ranked by similarity, with the time and camera for each.
    2. Suspect tracking: confirmed appearances across cameras are linked into a movement timeline.
    3. Real-time alerts: a department can load a watch list of specific, authorised individuals, such as wanted persons or missing children. The system flags a possible sighting to a control-room operator, who verifies it before anyone acts.

    The practical limit is the camera itself. A camera that is offline or badly placed cannot help anyone, so camera uptime matters as much as the AI.

    Your platform has supported over 2 lakh arrests, and you work with around 80,000 police officers across states. Can you walk us through one or two cases, such as tracking a suspect or identifying someone from footage, showing how the technology changed the outcome and how much faster it was than traditional methods?

    The platform has supported thousands of arrests across the country and is used by thousands of police officers. Many of these matters are still before the courts, so I’ll describe how cases typically unfold rather than name them.

    Tracing a suspect from a single frame. Take a street crime caught on one shop camera, where the investigator has just one frame of a face. Traditionally, officers collect footage from every recorder along likely routes and watch it, which can take days. Now they upload the frame and get back a short list of possible appearances across cameras, with times and locations. They verify each one, build the route, and move to the next lead, such as an address or a vehicle.

    Finding someone who has been missing for years. A person may have been absconding or missing for a long time, with only an old photograph on file. Matching that photograph gives officers a fresh lead on where the person has been seen. Officers then confirm identity through documents and witnesses before any action.

    The pattern is the same in both. Before, officers spent their time collecting and watching footage. Now they spend it checking a short list of leads.

    Recent protests have intensified the debate about facial recognition and the risk of it being used to monitor dissent or ordinary citizens. What do you say to concerns about mass surveillance, wrongful identification and bias?

    The concerns are legitimate, and anyone building this technology should take them seriously.

    NthEye is built to investigate specific crimes. It is not designed to watch lawful protests or track ordinary citizens. The safeguards are built into the product:

    1. Access controls:role-based access. Only authorised officers can run face searches.
    2. Purpose binding: a face search requires a case reference.
    3. Audit trails: every search is logged with who ran it, when and why, and the logs cannot be edited. Supervisors can review them.
    4. Retention limits: footage and search data are kept only as long as the department’s policy allows, then deleted.
    5. Human review: every match is a candidate with a score. A trained person reviews it, and the system never acts on its own.
    6. Bias testing: we test performance across age, gender and skin tone on Indian faces, and we flag low-quality images where confidence should be low.

    I would draw the line at purpose. Technology should help find a specific person connected to a specific crime, under an authorised, logged and reviewable process. It should not be used to profile everyone who walks past a camera.

    When a face match is wrong, who is accountable: the technology provider, the police or both? How do you make sure an AI match is treated as an investigative lead and not as proof? And given the absence of a comprehensive data protection and surveillance framework, how do you see regulation evolving?

    The provider and the police are both accountable, for different things. The provider answers for how the system behaves: tested accuracy, honest disclosure of its limits, clear confidence scores and tamper-proof logs. The police answer for what they do with a result.

    To keep a match a lead and not proof, the product never says “identified”. It says “possible match” and shows a score. Our onboarding trains officers to corroborate every match independently, through witnesses, records or physical evidence, before taking any action. A face match alone should never be the basis for an arrest.

    On regulation, the DPDP Act is now in force in phases. Its Rules were notified in November 2025, and most obligations apply from May 2027. But the Act gives the statewide room for law-enforcement processing, and India has no specific law on police use of facial recognition. I expect, and would welcome, clear national standards: accuracy testing before deployment, mandatory audit logs, rules on watch lists and retention, and independent oversight. Clear rules protect citizens, and they also protect responsible vendors from being lumped together with irresponsible ones.

    Through most of the 12 years, how was ValuePitch financed: bootstrapped, revenue-led, or with outside capital? What role will funding play in scaling this technology, and where will it go first: product, infrastructure, new states, or new use cases?

    ValuePitch has been fully bootstrapped from day one. Our verification and compliance businesses have funded the R&D behind every new product, including NthEye. That made us disciplined, because we only built what customers would pay for.

    As we scale, funding would go to four places, in this order:

    • Product: accuracy on low-quality Indian CCTV, processing at the edge, and making the safeguards stronger.
    • Infrastructure: the capacity to support state-scale deployments reliably.
    • Reach: more states, through local partners who can support police units on the ground.
    • New use cases: extending video intelligence to enterprises and commercial establishments.

    Explore more on TechTalk from ITmatters.

    1

    Leave a Reply

    Your email address will not be published. Required fields are marked *

    logo

    Get the latest news instantly

    You can change your preferences anytime.