metadata
title: Experimentlab
emoji: ⚛️
colorFrom: indigo
colorTo: purple
sdk: static
pinned: false
⚛️ Experimentlab (Experimentlab.in)
Sovereign Artificial Intelligence for Science, Technology, Engineering & Mathematics (STEM)
"Here’s to the rebels who build real things in the physical world. The ones who refuse to rent their intelligence by the token. Who pull the plug and keep calculating."
🌌 The Mission: Open Foundation STEM Intelligence
Experimentlab is dedicated to building sovereign, open-weights foundation models, efficient Small Language Models (SLMs), and offline workstations designed specifically for physical engineers, mathematicians, roboticists, and researchers.
We bridge deep academic theory with extreme computational efficiency:
- Zero Cloud Dependence: 100% offline, privacy-first execution on consumer laptops and edge devices.
- Mathematical Exactitude: Deterministic CAS (Computer Algebra System) verification for symbolic calculus, linear algebra, and discrete math.
- Sovereign Efficiency: Pioneering Sparse Mixture-of-Experts (MoE) and 1.58-bit ternary {-1, 0, +1} BitNet quantization running natively on CPUs and NPUs.
🏛️ Flagship Model Architectures
graph LR
A["Raw STEM Corpora (Ai2 Dolma, SymPy CAS, arXiv)"] --> B["Vigyan Sovereign Pretraining"]
B --> C["Vigyan MoE (1.5B - 3B 4x Experts)"]
B --> D["BitNet 1.58-Bit Ternary QAT"]
B --> E["Doctoral STEM Series (7B - 32B)"]
C --> F["Edge CPU & NPU (GGUF / bitnet.cpp)"]
D --> F
E --> G["Vigyan Pro Workstation"]
1. 🌟 Sparse Mixture of Experts (MoE)
Vigyan-1.5B-4x-MoE: Ultra-fast 4-expert routed SLM delivering 7B-class STEM reasoning at a fraction of inference memory.Vigyan-3B-4x-MoE: Advanced multi-expert router tuned on symbolic logic and scientific problem-solving.
2. ⚡ 1.58-Bit Ternary BitNet Architectures
vigyan-3b-1.58bit-qat: Pure 1.58-bit weights {-1, 0, +1} trained via Quantization-Aware Training (QAT). Eliminates power-hungry matrix multiplication in favor of integer addition, enabling offline execution on low-power ARM and x86 chips.
3. 🔬 Doctoral & Silicon Hardware Series
olmo2-7b-silicon-rtl-eda: Hardware silicon EDA specialist for Verilog RTL synthesis, timing constraints, and digital logic verification.olmo2-7b-phd-pure-math: Specialization in pure mathematics, formal proofs, topology, and abstract algebra.olmo2-7b-biomed-chem: Deep biochemistry, computational medicine, and molecular modeling.
💻 Vigyan Workstation: 100% Offline AI Workstation
All models pair seamlessly with Vigyan Workstation, featuring:
- Interactive WaveDrom Digital Timing Studio for electrical engineering and hardware timing diagrams.
- 6-DoF Robotics Kinematics Studio for forward and inverse DH-parameter manipulator simulation.
- Native SymPy CAS Mathematical Engine for symbolic calculus, derivatives, integrals, and LaTeX rendering.
- KiCad PCB Netlist Studio for local electronic circuit verification.
📂 Official Curated Collections
Explore our complete fleet grouped by domain:
- 🌟 Sovereign STEM SLMs & MoE Architectures — Flagship MoE, BitNet, and edge SLMs.
- 🔬 Doctoral Mathematics & Silicon EDA — Hardware silicon EDA, pure mathematics, astrophysics, and biomedicine.
- 📚 Open Datasets & Verified Research Whitepapers — Curated STEM pretraining corpora and research whitepapers.
Crafted with passion for sovereign open-source STEM research by Experimentlab.in • Built for the Crazy Ones.