SPEC-CDN Modular Setup v26.10
Copyright (C) 2026, Stream Platform Extrapolating Content Ltd
Initialising vector stream protocol ........ OK
Detecting business process ................. 1 found
Measuring flows ............................ OK
Measuring variance and cycle time .......... OK
Building statistical baseline .............. OK
Optimising before automating ............... OK
Loading modules: Forecasting, Vision, Legal data, Market risk ... OK
Press F10 to discuss a process, or scroll to enter SETUP
SPEC-CDN Setup Utility
System Information

We automate the optimised process, not the existing one.

Studio[SPEC-CDN]
Disciplines[Data science · Software architecture · Integration management]
Working with[Partners, not only clients]
Direction[Vector streams · content produced at the point of display]
[ Discuss a process ] [ Partner with us ] [ See the work ]

Method

Specification before code

Input[A business process as it actually runs]
Measure[Flows, variance, cycle time, dependencies]
Model[Statistical and econometric baseline]
Output[Architecture + automation, with before/after metrics]
  1. 1.Process diagnostics
    Statistical and econometric characteristics of the process: where it waits, where it varies, what drives the load.
  2. 2.Optimisation
    Redesign the process against the measured baseline before anything is automated.
  3. 3.Integration architecture
    How systems, data and protocols connect: a specification any team can build from.
  4. 4.Integration management
    Design and run the automation, with self-updating models and metrics before and after.

The order is the point. Automating a process before it is measured and optimised fixes its waste in software.

Partnership

Your process, our module

We look for partners, not only clients. A partner brings a real business process; we optimise and automate it, and the modules we build for it become part of our stack.

The partner gets[A working, measured improvement of its own process]
We get[A module proven on a real process]
Every module is[Applied first · reusable after]

We are not a corporation that funds a product line from its own budget. We build our stack through the processes of our partners, so every module solves a concrete problem before it becomes part of a product.

Direction

Stream the description. Extrapolate the content.

The company was founded to distribute content in new vector stream formats and protocols, built for augmented and virtual reality. With AI-generated media the case is stronger: send the description and the parameters, and let the content be produced where it is shown. That needs non-classical infrastructure and non-classical integration. Hence the CDN in the name.

In practice[AR elements in live broadcast at the right moment — branding the goal net when a goal is scored, for Vivacom Bulgaria]
Also[Dynamic advertising composed inside video streams]
Next[AI-generated media: the description travels, the content is produced at display]

Video is where this vision is most complete. It is not the only field: the data-science, forecasting and integration work below serves the same direction.

Work

From national energy forecasts to content in the stream and real-time market risk

► Forecasting · Self-updating forecast models

Statistical modelling software that forecasts electricity demand and supply for Bulgaria's Electricity System Operator. General linear models with composite predictors, updated automatically on a fixed cycle. The design anticipated tensor computation, so that many models can be refitted and many forecasts run at once.

Kostov, V., Tomov, L., Angelov, S. (2020). Software for statistical modeling via tensor algebra. Computer Science and Education in Computer Science, 16(1), 24–28. [CEEOL]

► Computer vision · Applied vision, built with start-ups
Goal detectionFrom video of sports pitches, the basis for automated match reports.
Augmented reality in live broadcastContent placed inside live video. See Direction.
Training analyticsStatistics extracted from training video of sports teams. Still in operation after several transformations.
Obstacle detectionA mobile application that warns visually impaired people of obstacles.
Camera analyticsMotion detection and object-type recognition on security cameras; free and occupied spaces in car parks.
Research visionSpecialised computer-vision development for the Institute of Metal Science, Bulgarian Academy of Sciences.
► Language and legal data · Classifying law at European scale

Work in academic projects on categorising national legislation against the EuroVoc thesaurus, including the MARCELL project. Our approach: algorithms that generate synthetic statistical corpora to stress-test document classifiers, together with neural networks and vector clustering.

► Now · Real-time market data and risk

Streaming pipelines from betting exchanges into our own databases, with order-book snapshots, settlement and derived risk metrics. On top of them, risk-analysis software and integrations for smaller sports-betting operators.

Contact

Bring us a process. We will bring the measurements.

A first conversation is about the process, not the software: what it does, where it hurts, what data it already leaves behind. We are a small studio, and each project gets a team assembled for it.

Host[host@spec-cdn.net]
EscTop F1Method F2Partnership F3Direction EnterNext F10Contact