| --- |
| title: ThinkNet |
| emoji: 🧠 |
| colorFrom: purple |
| colorTo: blue |
| sdk: gradio |
| sdk_version: 6.20.0 |
| python_version: 3.10 |
| app_file: app.py |
| pinned: false |
| --- |
| |
| # ThinkNet |
|
|
| ### Founded & Led by Naksh Gupta |
|
|
| **Building open-source AI systems and exploring the path toward intelligent, world-aware machines.** |
|
|
| ## 🌍 Our Vision |
|
|
| ThinkNet aims to explore and build the next generation of AI — systems that don't just generate text or recognize images, but can **understand the world, learn representations, predict consequences, reason about actions, and eventually interact with the physical world.** |
|
|
| Our long-term direction spans: |
|
|
| **LLMs → Multimodal AI → World Models → Embodied AI → Intelligent Agents** |
|
|
| ## 🧠 Our Current Stack |
|
|
| We work across: |
|
|
| * **Large Language Models** — Transformers, pre-training, fine-tuning, SFT & LoRA |
| * **Generative AI** — Diffusion Models & multimodal generation |
| * **Computer Vision** — CNNs, Vision Transformers, detection, segmentation & perception |
| * **World Models** — JEPA, latent-state prediction & action-conditioned prediction |
| * **Embodied AI** — perception, planning, control & robotics |
| * **AI Agents** — tool use, RAG, LangGraph & agentic systems |
| * **ML Engineering** — PyTorch, TensorFlow, Hugging Face, FastAPI & modern AI infrastructure |
|
|
| ## 🚀 What We're Working On |
|
|
| ### 🧠 ThinkNet LLM Series |
|
|
| Developing increasingly capable small language models, with a focus on **efficient training, high-quality data and open research**. |
|
|
| **Upcoming:** ThinkNet LLMs in the **100M–250M+ parameter range**. |
|
|
| ### 🌍 ThinkNet World Models |
|
|
| Researching **JEPA-style representation learning and world models** that can learn how the environment changes over time. |
|
|
| Current direction: |
|
|
| `Observation → Latent State → Action → Predicted Future State` |
|
|
| ### 🤖 ThinkNet Embodied AI |
|
|
| Exploring systems that connect perception with action: |
|
|
| `Vision → World Model → Planning → Control → Action` |
|
|
| The long-term goal is to build AI that can move beyond screens and **interact with the physical world.** |
|
|
| ### 👁️ Vision & Multimodal Models |
|
|
| Working toward models capable of understanding: |
|
|
| **Images • Video • Text • Audio • Physical environments** |
|
|
| ## 🔬 Upcoming Research |
|
|
| Our roadmap includes experiments and models around: |
|
|
| * **ThinkNet LLM** |
| * **ThinkNet Vision** |
| * **ThinkNet Multimodal** |
| * **ThinkNet World Model** |
| * **ThinkNet Embodied AI** |
| * **ThinkNet Agent** |
|
|
| These projects are developed incrementally through **open-source datasets, experiments, models and research implementations**. |
|
|
| --- |
|
|
| ## 🛠️ Our Philosophy |
|
|
| **Learn → Build → Experiment → Open Source → Iterate** |
|
|
| We believe meaningful AI progress comes not only from scaling models, but from discovering **better representations, learning objectives, architectures and ways for machines to understand and interact with the world.** |
|
|
| ### ThinkNet |
|
|
| **Building intelligence. Exploring the world. Open-sourcing the journey.** |
|
|
