---

📖 Table of Contents

• About Us
• Mission & Vision
• Flagship Project: Surya
• Model Portfolio
• Technical Specifications
• Architecture
• Getting Started
• Contributing
• Team
• Acknowledgments
• License

---

🏛️ About Us

SKT AI LABS is a frontier AI research and development organization headquartered in the heart of India. Founded and led by Shrijan Kumar Tiwari, we are a team of researchers, engineers, and data scientists committed to building sovereign artificial intelligence that serves the 1.4 billion people of Bharat.

Founded in: Sidhi, Madhya Pradesh & Raipur, Chhattisgarh, India

Our Philosophy

We believe that the future of AI must be inclusive, culturally grounded, and sovereign. While global AI models have made remarkable strides, they often fall short when it comes to understanding the linguistic diversity, cultural nuances, and practical needs of India. SKT AI LABS exists to bridge this gap.

Why Sovereign AI Matters

---

🎯 Mission & Vision

Mission

To develop world-class, sovereign AI systems that understand, respect, and serve the linguistic and cultural diversity of India, while maintaining the highest standards of data privacy and technological independence.

Vision

A future where every Indian developer, startup, and enterprise has access to frontier AI that speaks their language, understands their context, and operates within their control — without dependency on foreign technology.

Core Values

🛡️  SOVEREIGNTY   → Data and models under Indian control
🗣️  INCLUSIVITY   → Serving all languages, dialects, and communities
⚡  EFFICIENCY    → Optimized for accessible hardware
🤝  OPENNESS      → Open-sourcing key models for community benefit
🎯  EXCELLENCE    → Zero-hallucination, high-accuracy standards

---

🔥 Flagship Project: Surya

Project Surya is our ambitious initiative to develop India's own large-scale foundation models — built from the ground up for Indian contexts.

The Surya Series — Apex of Indian AI Performance

Target: 146 Trillion tokens database

---

🚀 Model Portfolio

Note: Most models are Closed Source. Only the experimental MoE variant was attempted to be open-sourced but the experiment did not succeed.

Released Models

SURYA-H IS AN EXPERIMENTAL MOE

In Development

Model Capabilities

┌─────────────────────────────────────────────────────────┐
│                    SKT AI LABS MODELS                   │
├─────────────────────────────────────────────────────────┤
│  📝 Text Generation    │  🖼️ Image Understanding       │
│  🎵 Audio Processing   │  🎬 Video Understanding       │
│  💻 Code Generation    │  🧠 Reasoning & Logic         │
│  🌐 Hinglish Mastery   │  📊 Data Analysis             │
└─────────────────────────────────────────────────────────┘

---

📊 Datasets

SKT-NRS — High-quality Indic dataset collection
Visit SKT-NRS on Hugging Face

---

⚙️ Technical Specifications

Model Architecture

Organization:      SKT AI LABS
Model Series:      SKT Surya Series
Architecture:      Custom Transformer-based (SKT_ST-X-0)
Specialization:    Advanced Coding, Logical Reasoning, Hinglish Mastery
Quantization:      GGUF, EXL2, AWQ (Broad Accessibility)

Performance Targets

• ✅ Hallucination Reduction — Fact-grounded responses
• ✅ 79+ tokens/s — Real-time inference capability
• ✅ Deep Hinglish — Slang, idioms, and dialect mastery
• ✅ Multi-Modal — Text, image, audio, and video understanding
• ✅ Energy Efficient — Optimized for accessible hardware

---

🏗️ Architecture

┌──────────────────────────────────────────────────────────────┐
│                    SKT AI LABS ARCHITECTURE                   │
├──────────────────────────────────────────────────────────────┤
│   ┌─────────────┐    ┌─────────────┐    ┌─────────────┐    │
│   │   INPUT     │───▶│  PROCESSING │───▶│   OUTPUT    │    │
│   │   LAYER     │    │   ENGINE    │    │   LAYER     │    │
│   └─────────────┘    └─────────────┘    └─────────────┘    │
│         │                   │                   │            │
│         ▼                   ▼                   ▼            │
│   Text/Image/Audio/Video  →  Custom SKT_ST-X-0 + Experts    │
│                                                               │
│   INDIAN CONTEXT MODULE (Hinglish, Culture, Values)         │
└──────────────────────────────────────────────────────────────┘

---

🚀 Getting Started

Prerequisites

# Python 3.10+
# CUDA 12.1+ (for GPU inference)
# 100GB+ RAM (for smaller models)
# 800GB-1TB GPU VRAM (for full-scale models)

---

🤝 Contributing

We welcome contributions from researchers, engineers, data contributors, and AI enthusiasts who share our vision of sovereign AI for India.

How to Contribute

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Open a Pull Request

---

📜 License

Our Models is licensed under the SKT Commercial License License.

SKT AI LABS Commercial License

---

📞 Connect With Us