Conversing with Streaming Data in Natural Language
An institution's data flows continuously. Transactions, logs, sensor outputs, and user actions accumulate every second. The real challenge is not storing this data, but understanding it while it flows and ensuring non-technical teams can also access it. Phexum Analytic Platform is a scalable data infrastructure that combines data analytics, application development, and enterprise insight generation under a single roof. This is also where Phexum's product journey began. This platform was our first major step, and independent products were subsequently born from within it.
Featured Results
The platform generates instant insight in high-volume systems by processing real-time data streams. Perhaps its most striking aspect was enabling natural language querying on streaming data even before LLM-based approaches became widespread. Users could generate reports with free text, even running location-based analyses using natural language. It offers developers the ability to develop and deploy applications within the platform using Python, JavaScript, and Java SDKs. Thanks to its distributed architecture, newly added servers automatically find each other to form a cluster, and all streaming data is distributed via load balancing for horizontal scaling.
Cross-Cutting Concepts and Architecture
The first concept running through this entire platform is accessibility. Data's value is directly proportional to who can reach it. That's why the platform decouples data access from technical expertise. Users can write queries and generate reports using free text, meaning natural language; they can even run location-based analyses without technical knowledge. The fact that it did this before LLMs became popular shows how accurate the approach was. The benefit is clear: working with data is no longer the job of a handful of experts, but of broader teams.
The second concept is scaling. It is impossible to process real-time, high-volume data with a single machine; therefore, we built the platform on a distributed architecture. Its technical name is the cluster approach. When a new instance is added to the system, this instance automatically discovers the others and joins a common structure; no one does manual configuration. The entirety of the streaming data is distributed over this cluster using load balancing. Its practical value is that the system can scale horizontally. When the data flow increases, you grow capacity by adding new servers, and performance is maintained even at high flow. On top of these two, a structure is added that gathers data coming from different sources into a single unified layer. Thus, scattered data turns into a centralized and consistent analytical model.
The World the Platform Lives In
The classic problem of data analytics is this: data sits in one place, teams try to access it from another, and technical people always build the bridge in between. A business team wanting a report is dependent on an expert to write the query. This creates both slowness and bottlenecks. On top of this comes the scale problem. As the data flow grows, a single machine is not enough. The system needs to spread across multiple servers while maintaining its consistency. Another layer is the scattered nature of data sources. An institution's data often sits divided across different systems. The Phexum Analytic Platform was built to address these three problems—the technical dependency of data access, the difficulty of scaling, and data scatter—under the same roof.
How the Story Was Built
This platform is the starting point of Phexum's productization story. As our first major step, we designed it with a focus on data analytics and enterprise insight generation, offering solutions to many companies on both software development and consultancy sides. At the center of its design, we placed not just analysis, but developability. The platform was not going to be a closed box consuming data, but an open system upon which applications could be written. This approach later paved the way for new products to be born from within it. The text-focused intelligence components separated over time and turned into an independent NLP platform.
Working with data is no longer the job of a handful of experts, but of the whole team.
Building Blocks of the Solution
The platform was built upon real-time stream processing. Applications can be developed and deployed inside it with Python, JavaScript, and Java SDKs; allowing solutions to be tailored to specific needs on the data. The natural language query layer makes it possible to generate reports and analysis with free text. The data consolidation layer gathers different sources into a single consistent model. The distributed architecture and automatic cluster structure enable horizontal scaling. Alongside these, the platform supports advanced analytical operations such as sentiment analysis, fraud/anomaly detection, and behavioral analyses. Together with the NLP platform operating integrated within it, it can also process unstructured text data under the same roof.
The Impact We Created
The Phexum Analytic Platform transformed scattered and continuously streaming data into a single analytical layer that both non-technical teams can access and that scales as volume grows. Institutions were able to contextualize data as it flows and turn it into instant insight, rather than merely accumulating it. For Phexum, this platform became both a product offering solutions to many companies and a fertile ground from which our subsequent products were born.