Why TRISHA Vision Exists
We started TRISHA Vision because India — with its languages, its complexity, and its scale — deserves AI that is built from the inside out. Most foundation models today are trained on Western data, fine-tuned for Western tasks, and translated for Indian users almost as an afterthought. The result is technology that speaks at India, not for it.
We are building something different. Foundation models trained on diverse Indian language corpora. Open datasets that the next generation of Indian builders can use freely. Applied AI tools where domain depth matters more than benchmark scores. And we are doing it openly — publishing what works, what didn’t, and how much it cost.
We are a small team. We work on consumer hardware. We move slowly when we should and fast when we can. Our mission is not to win an AI race. Our mission is to make sure Bharat shows up to the table with its own seat — and with something worth sharing.
Satya. Shakti. Shanti.
The three pillars our work is judged against.
Satya (Truth)
We don’t inflate claims. If we trained a 373M model, we don’t market it as a 7B model. If our corpus is 20 billion tokens, we don’t round it to 50. The credibility of Indian AI depends on the smaller players being more honest than the bigger ones. So we set our own bar high.
Shakti (Power)
Real capability, not theatre. We measure ourselves by what we can ship, what others can verify, and what real practitioners can use. Power without verification is just noise.
Shanti (Peace)
Sustainable building. We don’t burn out, don’t burn the planet, don’t burn community trust. Our infrastructure choices, our pricing, our pace — all reflect a long-term view. We are building for the next 30 years, not the next funding round.
What TRISHA Vision Is Becoming
TRISHA Vision is, at its heart, an attempt to build a Bharat-first technology ecosystem — grounded in wisdom, applied to real problems, built for everyone. AI is where we have the most depth today, and so AI is where we are starting. Foundation models, open datasets, and domain tools are our active focus.
Over time, we plan to extend this work into other technology areas where India lacks indigenous infrastructure — areas where being built in Bharat will matter for trust, for sovereignty, and for serving the actual needs of Indian users. We will share each direction publicly only when it is real and shippable, not before. That discipline is itself part of the mission.
For now: AI research. Done well. Done openly. Done for everyone.
How We Work
Open by default — Datasets, model cards, training notes, and lessons published freely. We monetize through enrichments, applied tools, and enterprise services — not by gatekeeping the foundation.
Practical over flashy — We use consumer hardware, optimize compute carefully, and publish honest cost numbers. Our work is reproducible by other indie builders.
Bharat-first, world-aware — Our defaults are Indian languages, Indian use-cases, Indian regulatory context. Everything we build also works for the wider world, but the order of priority is clear.
Long-horizon — We are not optimizing for the next 12 months. We are optimizing for the kind of organization that will still be useful in 2050.
Built in Bharat 🇮🇳 — ek behtar duniya ke liye.
Have a question, a collaboration idea, or a use case to share? We’d love to hear from you. → Contact us