AI systems today can explain, summarise, and predict, but they still tend to approach problems one way at a time. This is unlike human thinking, which draws from multiple perspectives. That gap is what Raja Dharma Tej Maddala, a Class VII student from Hyderabad, has been working on, by trying to separate different modes of reasoning instead of letting a system follow a single path.
“Raja MagRex AI is already a working product at the prototype stage, not just a concept. The current system includes a functional backend, a structured interface, and the ability to operate through defined systems and personas that influence how responses are generated,” he says.
In its current form, the system lets a user approach the same problem through different reasoning modes, such as logic, creativity, emotional understanding, or long-term analysis. Personas then affect how the answer is framed within that mode. Most significantly, the user remains the one making those choices. Raja wants the system, over time, to decide more of that on its own.
That also marks the boundary of the project. This is not a new foundational model. It is a structured reasoning interface built through system design and prompt-led organisation. Raja does not claim it is better than conventional AI models across the board. “At this stage, I do not claim that Raja MagRex AI is better than conventional AI models. My testing approach focuses on evaluating differences in structure and reasoning rather than raw performance metrics,” he says. Formal benchmarking is still to come.
The system is not positioned as a general-purpose solution. Its current range reflects both design choices and the limits of what has been implemented so far. “Raja MagRex AI performs best in structured explanation and idea exploration, but it is not yet designed for highly critical or high-stakes decision-making. It also does not yet maintain long-term memory across sessions, which limits continuity in complex workflows,” he says.
Part of what makes the project worth tracking is the way Raja describes his own thinking. He links the system to both AI and physics, especially to ideas of complex systems, consistency, and constraints. He also speaks less like someone trying to prove a grand theory than someone working through engineering trade-offs. He says the hardest part was reducing a large architectural idea into something stable enough to test.
“When I approach a problem, I begin by identifying what the problem is asking, what type of reasoning it requires, and what constraints are involved. One important shift in my thinking was moving away from the idea that there is always a single correct way to solve a problem,” he smiles.
Raja MagRex AI is a prototype in progress, not a finished system and not a claim of technical disruption. The more interesting development here, perhaps, is that a school student is already thinking about AI at the level of reasoning design, not just using it to finish homework.
