File size: 2,511 Bytes
00bfe7c cf91278 00bfe7c cf91278 33ef360 cf91278 03bff2e cf91278 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 | ---
organization: AMFORGE
---
<div align="center">
# AMEFORGE
**Independent AI Research Studio**
[Website](https://ameforge.tech) · [GitHub](https://github.com/Volgat) · [Google Play](https://play.google.com/store/apps/dev?id=8746471888242300768)
</div>
---
## About
Ameforge is an independent AI research studio. We design novel architectures, publish peer-reviewed research, and deploy production-ready models with software that brings these innovations directly to users.
Our work spans deep learning architecture research, natural language processing, time-series forecasting, and mobile software development.
---
## Research
### EARCP — Ensemble Adaptive Recurrent Cascade Predictor
Published on arXiv · Validated on Kaggle Hull Tactical Market Prediction (Rank 708 / 18,000+ teams · Top 24%)
A multi-expert ensemble architecture combining CNN, BiLSTM, and Transformer components for adaptive time-series prediction. Applicable to financial forecasting, sensor data, computer vision, and autonomous systems.
```bash
pip install earcp
```
📄 [arXiv Paper](https://arxiv.org/abs/2603.14651) · 🏆 [Kaggle Leaderboard](https://www.kaggle.com/competitions/hull-tactical-market-prediction/leaderboard)
---
### SparseMind — Sparse Neural Architecture Research
Experimental research into high-sparsity neural architectures achieving 87.5% weight sparsity while maintaining competitive perplexity scores. Designed for efficient inference on resource-constrained environments.
---
### NexusBPE — Custom Tokenizer Pipeline
A custom Byte-Pair Encoding tokenizer trained on a curated multilingual corpus (code, natural language, technical documents). Designed to complement Ameforge's internal language model research.
---
### MemoryBank (LPOLMemory) — Continual Learning Research
Research into long-term memory mechanisms for neural networks, addressing catastrophic forgetting in continual learning scenarios.
---
## Models
All published models are available under the **CKL (Custom Knowledge License)**. Commercial licensing available — contact [contact@ameforge.tech](mailto:contact@ameforge.tech).
| Model | Task | Status |
|---|---|---|
| gc_editor1 | Text Generation | Active |
| gc_editor | Text Generation | Active |
| gearcut_tok | Tokenizer | Active |
---
## Contact
📧 [contact@ameforge.tech](mailto:contact@ameforge.tech)
🌐 [ameforge.tech](https://ameforge.tech)
---
<div align="center">
<sub>© 2025 Ameforge. All rights reserved.</sub>
</div>
|