Sitemap
Medium Logo

I Don't Wait for the Tools to Exist

There is a specific kind of frustration that does not have a name in most research departments. It is the frustration of building a system that works by every available measure and knowing, with a certainty you cannot yet formalize, that it is still wrong.

Not wrong in the way a bug is wrong. Wrong in the way a question is wrong. Wrong because the instrument you are using to measure it was never designed to see what you are looking for.

I have spent most of my research life in that gap. The gap between what the field currently measures and what actually matters. The gap between what a system reports and what it does. The gap between a correct answer and a correct answer in time.

My name is Prakul Sunil Hiremath. I am an independent researcher, systems thinker, and the founder of Aliens on Earth — a research collective built from scratch in Belagavi, Karnataka, India. I am also a B.Tech Computer Science student at Visvesvaraya Technological University, and I offer that fact not as an apology for what follows, but as the first piece of evidence for it.

The institutional address has never determined the gravity of the question.

Building in the Mud

There is a version of research that happens inside clean, well-funded rooms with established pipelines and pre-approved methodologies. That version has produced important work. It has also produced an entire ecosystem of problems no one inside it is structurally incentivized to name.

I do not work in those rooms. I work in the trench: the place where you cannot cite a tool that solves your problem because the tool does not exist yet, where the path from question to answer runs through architecture decisions at two in the morning, where the only thing that stops you from quitting is the fact that the problem is real and the gap is not theoretical.

The absence of institutional infrastructure is not a handicap I am managing. It is a methodology. When you cannot appeal to a department's existing framework, you are forced to think from first principles. You are forced to ask not just "what is the answer" but "what is the right question" and those are not the same question.

I have found, consistently, that the most important work lives in the space between existing disciplines. The mathematician does not talk to the security engineer. The machine learning researcher does not understand what is actually happening inside a live biological reactor. The astrophysicist and the anomaly detection expert are solving the same fundamental problem; how do you detect a structural state transition in a high-dimensional signal stream before the window closes and they have never been in the same room.

That gap is not inevitable. It is institutional. And it is exactly where I choose to operate.

Aliens on Earth

Every significant project I have built has started with the same recognition: the problem I am trying to solve does not fit inside any single department's jurisdiction. It lives at a boundary that existing structures were not designed to cross.

Aliens on Earth — AoE — exists because of that recognition.

AoE is not a company. It is not a lab. It is an agile research collective built on a single thesis: the mathematical laws that govern high-stakes systems do not care about university departments, and the people who want to understand those laws should not be stopped by them either. AoE is the structure I built to allow that kind of thinking to happen in practice — to bring together the right combination of expertise, curiosity, and willingness to sit with hard problems long enough to actually solve them.

The name is deliberate. We are interested in the things that do not fit the existing map. The questions that make specialists in adjacent fields say "that is not my domain." The problems that require you to be, simultaneously, a mathematician, an engineer, a biologist, and a systems architect — and to hold all of that at once without flinching.

That is the kind of thinking AoE was built to protect.

What I Am Actually Obsessed With

I am not interested in AI as a technology trend. I am interested in a specific class of problem that I have encountered in every domain I have worked in, from cybersecurity to astrophysics to climate infrastructure:

Systems that fail silently. Before their creators know the window has closed.

This is the problem that unifies everything I build. A cyber-attack that has already restructured the network before a single alert fires. An anomaly detection model that scores perfectly on every benchmark while failing at the only thing that operationally matters: timing. An algae reactor running outside safe parameters while every sensor reading stays within nominal range. An astrophysical transient that has already decayed before the detection pipeline processes the observation.

In every case, the system is not failing loudly. It is failing in the gap between what the instruments measure and what the system is actually doing. It is failing because the evaluation framework was designed for a static world and deployed into a dynamic one.

My obsession is with the instruments themselves. With the question of what it means to actually measure a system that operates in time, under uncertainty, at the boundary of its own stability. With building the frameworks that make invisible failures visible — not after the fact, but in the window where intervention is still possible.

This has led me into sequential machine learning, probabilistic state-space models, temporal evaluation theory, bio-conformal architectures for biological systems, and uncertainty-aware signal analysis across multi-wavelength astrophysical observations. These are not separate interests. They are the same obsession applied to different domains where the cost of a late detection is real and irreversible.

The Art of Translation

Independent research at the boundary of multiple disciplines requires a skill that most academic training does not develop: the ability to make a technically rigorous idea legible to someone who did not build it.

I think of this as translation. Not simplification — the mathematics does not get simpler because the audience is non-specialist. But the way you enter a problem, the language you use to name it, the analogy you choose to make the structure visible — these are not decorative choices. They are epistemological ones. The right translation does not reduce an idea. It reveals its structure to someone who was previously unable to see it.

I have had to develop this skill by necessity. Presenting at IIT Bombay and ISRO's AstroSat Conference. At the Asian Aerosol Conference in Mumbai. At the India AI Impact Summit at Bharat Mandapam in New Delhi — national-level rooms where the audience spans disciplines and the ideas have to hold under scrutiny from every direction simultaneously. In government school classrooms across Karnataka, where the challenge is not simplification but ignition: making a student who has never heard of machine learning understand why a question about how systems fail is one of the most interesting questions in the world.

That range is not incidental. It is the whole point. The ideas I care about are important enough to reach every room they need to reach. The work of translation is not separate from the work of research. It is the final test of whether you actually understand what you built.

On Writing

I write because the gap between what the field currently discusses and what actually matters in production systems is wider than most practitioners realize — and naming a gap clearly is the first step toward closing it.

I do not write tutorials. I do not write surveys of tools other people built. I write about the moments when the existing framework revealed its own limits: when two models with identical AUROC scores turned out to be operationally non-equivalent, when a perfectly calibrated system failed at the only variable that mattered, when the instrument designed to measure the thing turned out to be blind to the thing itself.

Those moments are where the real questions live. And they are the questions I intend to keep writing about for as long as the field keeps pretending they do not exist.

The Standard You Build When the Existing World Fails to Measure Up

Every framework I have built — every metric, every architecture, every paper that starts from first principles rather than from prior art — began the same way. Not with an idea, but with a failure. Not with ambition, but with a problem that the existing tools could not solve and that I could not walk away from.

The HED Score did not begin as a research agenda. It began as a wall I hit in the middle of a detection project where every number looked correct and the system was still failing. The wall was the realization that the instrument I was using to evaluate the system was constitutively incapable of seeing the failure mode I was looking at. Not because I was using it wrong. Because it was not designed to look at time.

You do not write a paper about that realization. You build a standard.

That is the principle I operate on. Not because institutional inertia is something to be celebrated, but because some problems will not wait for the right department to authorize the right research program. Because the gap between what is currently measured and what actually matters is a real gap in the real world, with real consequences for real systems. And because the only answer to a gap like that is someone willing to stand in it and build something new.

I am an undergraduate at a state university in Karnataka. I have no corporate budget and no institutional brand. I have a research collective, a set of first principles, and a record of frameworks built because the existing ones were not sufficient.

That is enough. It always has been.

📬 prakulhiremath03@gmail.com 🌐 aliensonearth.in/prakulsunilhiremath 📦 github.com/prakulhiremath