Your Astrology Business

This article is part of my “Your Astrology Business” category series, where I explore setting up and running a small astrology business for yourself.

Guest Post

By Abhishek Gupta from KunliAI

What Building an AI That Reads Kundlis Taught Me About Astrology I Never Knew

I didn’t grow up believing in astrology. My mother would occasionally mention someone’s Mahadasha or ask about a Muhurat before a big decision, and I’d nod along without really absorbing what any of it meant.

So when I decided to build an AI platform that generates Vedic astrology readings, I assumed the hard part would be the engineering — the AI, the calculations, the software. I was wrong.

The hard part was astrology itself.

The Engineering Wasn’t the Hard Part

I came into this project the way most people with a technical background do: assuming the domain knowledge would be the easy part and the software would be the challenge. Get the birth chart calculations right, connect a language model, format the output nicely — how hard could it be?

The calculations, as it turned out, were the easy part. Swiss Ephemeris data is precise, well-documented, and mathematically unambiguous. A person’s planetary positions at the moment of their birth are a solved problem. What is not a solved problem is knowing what to do with that data once you have it — and that gap is where I spent the vast majority of my time.

Why Vague Astrology Sounds Astrological Without Being Astrological

The first thing I learned is that astrology has almost no room for vagueness, even though from the outside it can look like it’s made entirely of vague statements. Every placement means something specific, and that specificity has to connect logically to something else — a house, a planet’s dignity, a dasha period.

When I first tried feeding basic birth chart data into an AI system and asking it to write a “personalized reading,” it produced confident, well-written nonsense. It would say a person’s Mars in the seventh house made them “passionate in relationships,” which sounds plausible until you actually understand what Mars in the seventh house does to marital timing, temperament, and the specific tensions it creates — and none of that showed up in the generic paragraph the system wrote. It read like astrology. It used the right vocabulary. But it wasn’t actually derived from the chart in front of it — it was the kind of sentence that could have been written about almost anyone.

That failure taught me something I didn’t expect: astrology’s credibility with people who actually know it doesn’t come from the predictions being dramatic. It comes from the reasoning being traceable. A good astrologer can tell you why they’re saying something — which planet, which house, which dasha — and a bad reading is one where you can’t trace the “why” back to the chart at all.

The Principle That Rebuilt the System: Traceable Reasoning

Once I understood that, I ended up rebuilding the entire system around one principle: every statement had to be explicitly tied to a specific placement, written as “because this planet is here, this happens,” rather than a free-floating description that sounded astrological but wasn’t actually derived from anything.

This sounds simple to state and turned out to be genuinely difficult to enforce, especially once I started working across languages. I built the platform to serve Hindi-speaking and Tamil-speaking users first, not English-speaking ones, and I quickly discovered that the “because X, therefore Y” logic doesn’t survive translation on its own. It’s easy to lose the causal chain in translation — you end up with a sentence that’s grammatically correct in Hindi but has quietly dropped the reasoning that justified it in the first place. I had to build explicit checks into the system just to make sure a prediction’s stated cause and its stated effect actually stayed connected to each other in every language the platform supported.

It also meant getting strict about remedies, which I hadn’t expected to be a source of errors at all. Early on, the remedy suggestions the system generated were generic — a mantra here, a gemstone there — without being tied to the specific planet governing a person’s current dasha period. A remedy that isn’t connected to the actual planetary period someone is running through isn’t really a remedy in any traditional sense; it’s decoration. I ended up building an explicit mapping between each graha and its correct mantra, gemstone, and presiding deity, so that what the system recommended was actually anchored to the person’s chart rather than pulled from a generic list.

What I Never Understood About Nakshatras

The second thing that surprised me was how much nuance lives in nakshatras — something I’d genuinely never paid attention to before this project. I’d assumed the moon sign was the main event, the way most casual Western astrology treats a sun sign. But nakshatras carry an entire layer of mythology, ruling deities, and temperament that shapes a reading in a way the moon sign alone doesn’t capture.

Building a system that had to explain a nakshatra’s meaning to someone with zero background meant I had to actually understand the mythology behind it — why a particular nakshatra’s ruling deity matters to the interpretation, not just recite that it does. This turned into one of the more humbling parts of the project. I could look up a nakshatra’s name and its broad characteristics easily enough. Understanding why those characteristics follow from the deity and the symbolism attached to it — understanding it well enough to explain it correctly to someone else, in their own language, without flattening it into a generic personality description — took considerably longer, and I still don’t think I’ve fully arrived at it.

Dashas Taught Me That Timing Isn’t a Detail — It’s the Whole Point

If there’s one area where I underestimated the complexity going in, it was dashas. I understood, abstractly, that Vedic astrology places enormous weight on timing — that the same planetary placement can mean something different depending on which period of a person’s life it’s activated in. What I didn’t appreciate was how much of what people actually want to know from a reading is a timing question, not a character question.

People don’t come to a Kundli reading mainly to be told what kind of person they are. They come to ask when — when will there be movement in a stalled career, when is a good period for marriage, when does a difficult phase end. That reframed the entire project for me. Getting a system to correctly identify a person’s current Mahadasha and Antardasha, and to speak to what that specific combination of periods tends to bring, mattered more to the people using the platform than almost any other single feature I built. It was also one of the places where getting it wrong is most obvious to anyone who actually knows their own chart — there’s no room to be approximately right about which dasha someone is in.

What I Came Away With

What I came away with wasn’t belief in the predictive power of astrology in some mystical sense — that’s a separate question, and not one I’m trying to settle here. What I came away with was real respect for astrology as a system. It has internal logic. It has rules that don’t bend just because a generic, feel-good sentence would be easier to write. Trying to get a machine to follow that logic honestly, instead of faking it, turned out to be the most humbling part of building anything I’ve built.

It also changed how I read astrology content myself, outside of my own project. I notice now when a reading is actually reasoning from a chart versus when it’s assembling astrology-flavored language around a conclusion it would have reached anyway. I don’t think I would have been able to tell the difference before I had to build a system that was forced to show its work.

If there’s a lesson in this for anyone approaching astrology from the outside, skeptical or curious, it’s this: the depth is in the specifics, not the generalities. Anyone can write “you’re going through a transformative period.” Far fewer people — or systems — can tell you why, and have that “why” hold up when you check it against the chart.

Author Bio

Abhishek Gupta is the founder of KundliAI, an AI-powered Vedic astrology platform built for Hindi and Tamil-speaking audiences. He began building the platform as a solo founder in 2026 with no prior astrology background, learning the discipline from the ground up in order to build a system that reasons from a chart rather than generating generic predictions.