Kepler Systems

Building the future of AI through collaborative research, shared datasets, and open-source models. | Managed by @XCossale

[Kepler Systems @KeplerSystems]

1/5 Launching Kepler Systems—an AI research initiative where I build open models and share datasets. Starting with datasets I’ve curated to preserve Indo-Pak Urdu poetry’s richness. No corporate labs here—just a solo researcher and a lot of love for language and art.

[Kepler Systems @KeplerSystems]

Technical details & samples: https://t.co/uRV8wmda3t

Datasets: https://t.co/X8jg7Vmzq3

Support PAARI: https://t.co/nRDhkT8kdZ

Share if useful for your research.

[Kepler Systems @KeplerSystems]

We've released 8 multilingual datasets from rural India.

19.9M tokens across Hindi, Urdu, Punjabi, Tamil, Telugu, Marathi, Gujarati, and English.

Indian languages represent 1B+ speakers, but quality training data is scarce. These datasets help address that.

🧵 Thread

[Kepler Systems @KeplerSystems]

From PAARI: "All our content is, and will forever be free to consume."

Please cite PAARI and this dataset when using. Respect journalism attribution.

Building AI for Indian languages requires data that represents real experiences.

[Kepler Systems @KeplerSystems]

This is just the start. I’m exploring other things —no hype, just incremental progress. Follow along or reach out at [email protected]. Let’s make AI silly and dumb.

Explore the work: https://t.co/nO0is2DWgb
Built with ♥️ in 🇮🇳 by @XCossale

[Kepler Systems @KeplerSystems]

Building on these datasets: Poetry Llama, a 70B-parameter model fine-tuned to understand and generate Urdu poetry.
- Trained on 7M tokens
- Llama license for research/commercial use
- Experiment with it → https://t.co/0mIzuNdZnp

[Kepler Systems @KeplerSystems]

UrduGhazals-25k: 25,000 complete ghazals spanning dialects and eras.
- Designed to support multilingual research
- Free to use (CC BY-SA 4.0)
- Explore → https://t.co/VtIV6O01UD

[Kepler Systems @KeplerSystems]

UrduShers-10k: A collection of 10,000 Urdu couplets.
- Sourced from classical to modern poets
- Quality-validated for consistency
- CC BY-SA 4.0 license
- For poets, linguists, or AI enthusiasts → https://t.co/ig9fiZa36o

[Kepler Systems @KeplerSystems]

1/5 Launching Kepler Systems—an AI research initiative where I build open models and share datasets. Starting with datasets I’ve curated to preserve Indo-Pak Urdu poetry’s richness. No corporate labs here—just a solo researcher and a lot of love for language and art.