Independent Research & Development
EONFLUX

Researching intelligent systems, artificial intelligence and computational engineering.

Scroll↓

Research

Experimental research in artificial intelligence, learning systems and computational engineering.

01

Direct Dataset Solving

Research into mathematical methods for deriving a function directly from noisy labeled observations, rather than discovering its parameters through iterative optimization. Initial experiments with linear datasets produced solutions comparable to gradient descent, with early polynomial extensions showing similar behavior. The research also revealed that data ordering influences the resulting solution, opening questions about deterministic solution spaces and informed initialization for iterative learning.

02

EON Project — Evolving Ontogenic Network

A neural architecture designed to grow as it learns, expanding only when new knowledge requires additional capacity while preserving previously acquired representations.

03

Program Synthesis from Function Libraries

Learning ordered compositions of available functions that transform a given input into a desired output, discovering valid and sometimes more efficient computational paths automatically.

04

LLM Hardware & Workload Modeling

Modeling the relationship between model size, quantization, context length, concurrent sessions and hardware constraints to determine viable deployments — or the best model for a given machine.

Next ↓Community

Community

Open tools and infrastructure developed by Eonflux for developers, researchers and the broader technical community.

Pyon — Python Object Notation

An open serialization format and Python library that extends JSON with native support for complex Python types, preserving structure and type information across serialization and reconstruction. Designed for advanced data workflows involving AI, machine learning, NumPy, pandas and custom Python objects.

View on GitHub ↗

MTLink

An open multi-terminal communication layer for MetaTrader environments, designed to coordinate multiple terminal instances and broker accounts through a unified, typed interface.

GitHub — Coming Soon
Next ↓About

About Eonflux

Eonflux is an independent research and development initiative focused on artificial intelligence, learning systems, computational engineering and open technical infrastructure. Our work explores both foundational research questions and practical systems, with an emphasis on building methods, architectures and tools that can be tested, measured and evolved.

Research first. Engineering when it matters.

Next ↓Contact