Naré Vandanyan, Co-Founder & CEO of Ntropy, a platform that enables developers to parse financial transactions in under 100ms with super-human accuracy, unlocking the path to a new generation of autonomous finance, powering products and services that have never before been possible. It converts raw streams of transactions into contextualized, structured information by combining data from multiple sources, including natural language models, search engines, internal databases, external APIs, and existing transaction data from across our network.
You grew up in Armenia, without electricity during a war. Could you share some details regarding these early days, and how this led you to work for the United Nations?
That experience was shared by an entire generation in Armenia. It fostered in me a sense of imagination and the ability to find solutions even with little means. Like others who grew up in a conflict zone, this period in my life had a profound impact on how I see the world. These demanding circumstances nurtured a sense of shared responsibility within the community and a resolute drive to bring about positive change. Realizing that our challenges extended beyond individual struggles, I felt a calling to think on a broader scale and channel my endeavors. This, in turn, steered me towards the United Nations.
The UN emerged as the ideal platform to contribute meaningfully. Given Armenia’s precarious geopolitical position and my aspiration to influence global matters, I believed that collaborating with the UN would offer an opportunity to truly make a difference. By being part of consequential discussions and decisions, I aimed to have a meaningful impact on the world’s issues.
You soon became disillusioned with the United Nations, how did you then shift to wanting to work in tech?
The disillusionment with the UN was rooted in its slow and bureaucratic nature, which eventually prompted a shift in my career aspirations. While the UN had its advantages, I came to realize that it often lacked effective action and the ability to drive authentic change. This realization guided me to redirect my focus toward the realm of technology – a dynamic and unrestrictive space.
In the world of technology, innovative tools are readily available and constantly advancing, granting individuals the ability to spark transformation without unnecessary hurdles. This environment fosters the transformation of ideas into reality, unhindered by unnecessary permissions – a facet that really fascinated me. The potential to make a substantial, widespread impact through technology became an irresistible calling, compelling me to immerse myself in this vibrant field.
What were some of the first data projects that you worked on?
One of my earlier projects was creating an app focused on teenage mental health. The app used passive haptics data and conversational intelligence to identify early signs of bipolar disorder. At that time, the field of natural language processing was not as advanced as it is today, which is quite remarkable considering it was only about six years ago when this project was initiated. Our work was one of the first research and development initiatives in this space, and we later sold our IP to insurers for internal analytics and underwriting.
You previously invested in AI and ML companies through the London-based AI Seed, what were some of the common traits that you observed with successful AI startups?
A constant thread was having exclusive access to data, along with the ability to harness this data to tackle real-world problems. Moreover, it’s crucial to acknowledge that within the realm of applied AI companies, the emphasis goes beyond just constructing models; it shifts towards creating impactful, valuable products. Teams that grasp and embrace this viewpoint are the ones that genuinely thrive in the AI/ML landscape. For example, Predina uses AI to predict the risk of a vehicle accident for a given location and time, while Observe Technologies uses proprietary algorithms to support fish farms to sustainably grow food.
Could you share the genesis story behind Ntropy?
Ntropy was born out of the idea that some of the world’s most important information is hidden in financial transactions. Until now, this data has lived in silos, which is messy and difficult to work with. We created Ntropy to be the first truly global, cross-industry, cross-geo, and multilingual financial data engine that can provide human-level accuracy. By creating a common language and system to understand financial data, we are equalizing trust and access to money for businesses and individuals anywhere. By having the ability to understand and interpret these transactions, the dynamics of money can be redefined, along with accessibility to it.
We’ve had quite the archetypal startup story. In the beginning, my co-founder Ilia and I were operating from an abandoned dusty school building basement. We started with 20k transactions and a distilled BERT model trained on them. The data was bootstrapped from a consumer app on Typeform with a Plaid connection, and supported by friends and family. We were working long hours and strapped for cash in the beginning, but fueled by determination and dedication to this business.
Fast forward to today, our journey has led us to analyze and label billions of transactions. As a result, we now have one of the world’s most comprehensive merchant databases with close to 100M+ merchants enriched with names, addresses, industry tags, and more. We’ve consistently expanded our repository of transactions – harnessing the power of LLMs on this financial data has delivered unparalleled cost-efficiency and speed. This capability holds the potential to revolutionize the financial landscape.
Why is financial data one of the great equalizers?
Financial data emerges as a powerful equalizer due to its capacity to level the playing field, reduce uncertainty, and foster trust. When data is abundant and refined, it translates to diminished risks linked with financial decision-making. As risk becomes more manageable, a shift happens. The cost of uncertainty diminishes, enabling individuals to make more informed and equitable decisions, which in turn levels the playing field. For example, if we have greater access to data and no longer make decisions based on a very narrow set of parameters, a new immigrant has the same potential as someone from a well-established lineage to secure favorable terms on a car loan or mortgage. Essentially, the obstacle presented by financial imbalances begins to dissolve, introducing an era where a wider range of people can access advantageous financial opportunities.
What are some of the challenges behind building an AI that can read and understand financial transactions like a human would?
Developing AI capable of comprehending financial transactions like humans can is challenging due to its probabilistic nature, which can lead to errors. Unlike humans, AI systems still lack accountability structures. The main challenge is refining AI systems to reduce errors and their impact while ensuring scalability. Interestingly, larger models can alleviate this challenge by gradually improving accuracy over time. Amplified capabilities and a wealth of data can enhance AI’s interpretive accuracy, ultimately cultivating a more lenient error-tolerant environment and expediting the widespread adoption of these systems.
Can you discuss how Ntropy offers standardized financial data?
Ntropy functions as an all-encompassing platform, bringing together a spectrum of language models, spanning from the most extensive to the most compact, in conjunction with heuristics. These models are trained using raw financial data, expert insights, and machine-labeled samples. Our goal is to extract meaningful insights from a variety of transaction strings and present them cohesively in an easily understandable way. Our suite comprises APIs and an intuitive dashboard, enabling the rapid conversion of financial data within milliseconds. This functionality seamlessly integrates into users’ products and services.
What are some of the use cases behind this data?
The applications for this data are extensive, spanning the entirety of financial operations. It empowers diverse functions including payments, underwriting, accounting, investing, and more. The adaptability of the data becomes clear in its ability to impact various aspects of financial activities, whether it involves fund transfers, meticulous record-keeping, or optimizing capital utilization.
Consider bank transactions or a budgeting app. A quick look reveals the difficulties in understanding purchases due to non-standard merchant names and descriptions. While many companies have attempted to address this issue through internal solutions, they often fall short in terms of scalability, maintenance, and generalization. A custom model is generally only 60-70% accurate and can take months to build.
Ntropy’s technology combines billions of data points from global merchant databases, search engines, and language models trained on a condensed version of the web to process banking data across four different continents and six-plus different languages. We’re enabling the use of large language models at scale in finance to support all back-office functions.
What is your vision for the future of Ntropy?
Our vision for Ntropy is clear: We aim to become the go-to Vertical AI company for financial services. Our strong foundation of data and intuition, supported by a dedicated team, has uniquely positioned us to drive real change. So, what does this actually mean in practice? It’s about leveraging the latest advancements to transform finance and unlock new levels of productivity that were previously out of reach.
We all know banking can be expensive. But imagine if we could change that. By reducing costs, we’re not just cutting expenses, we’re encouraging healthy competition, improving the economics of the system, and ultimately making financial services more accessible and efficient for everyone. That’s the future we’re working towards – a financial landscape that’s fairer and more user-friendly.
Thank you for the great interview, readers who wish to learn more should visit Ntropy.
Credit: Source link
