README / README.md
Keyven's picture
Karte mit Standard-Überschriften; Space-Seite hell mit dunkler Schrift
c41989f verified
|
Raw History Blame Contribute Delete
4.26 kB
---
title: Keyvan.ai
emoji: 🧩
colorFrom: blue
colorTo: indigo
sdk: static
pinned: false
---
![Keyvan.ai](https://huggingface.co/spaces/keyvan-ai/README/resolve/main/banner.png)
**Keyvan.ai** is an independent AI research and engineering practice based in Germany (Keyvan Hardani). We build and evaluate language, vision, audio and security models for production use, with a focus on small models that run on-premises, and we publish our experiments with full protocols.
## Current research: weight-space programming
Editing the weights of neural networks directly with a frontier model as the agent, without gradients or training, verified by a fixed evaluation harness. Following the MIT CSAIL study *Sparks of Weight-Space Programming* (2 November 2026), we replicated and extended the approach with Claude Mythos 5.1.
- [Mythos-WSP-Qwen3.5-0.8B](https://huggingface.co/keyvan-ai/Mythos-WSP-Qwen3.5-0.8B) – Qwen3.5-0.8B with two weight edits: a rank-3 projection and an implanted repetition-detection circuit in four repurposed MLP units. MATH-500 (thinking mode, 32k tokens): 7.3 % → 58.7 %, truncated answers 93 % → 17 %. No training.
- [Mythos-WSP-Loopbreaker-Skill](https://huggingface.co/keyvan-ai/Mythos-WSP-Loopbreaker-Skill) – the implanted circuit as a separate, versioned tensor set with an application script.
- [Research report](https://doi.org/10.5281/zenodo.24036187) – *Weight-Space Programming with Claude Mythos* (bilingual EN/DE), DOI 10.5281/zenodo.24036187, including a transfer of the method to a thermal surrogate model from power electronics.
![MATH-500](https://huggingface.co/spaces/keyvan-ai/README/resolve/main/abb1_math500.png)
## Models
**German language (Mankei family)**
- [Mankei-1B-Chat](https://huggingface.co/keyvan-ai/Mankei-1B-Chat) · [Mankei-1B-Chat-GGUF](https://huggingface.co/keyvan-ai/Mankei-1B-Chat-GGUF) – 1B German chat model, safetensors and GGUF
- [Mankei-Boarisch](https://huggingface.co/keyvan-ai/Mankei-Boarisch) – 1B Bavarian-dialect text generation
- [Mankei-326M-Embedder](https://huggingface.co/keyvan-ai/Mankei-326M-Embedder) · [Mankei-326M-Reranker](https://huggingface.co/keyvan-ai/Mankei-326M-Reranker) – German retrieval: sentence embeddings and reranking
- [Mankei-Reflex](https://huggingface.co/keyvan-ai/Mankei-Reflex) – text classification
- [german-text-3.1](https://huggingface.co/keyvan-ai/german-text-3.1) – 2B German text generation
**German document understanding (German-OCR family)**
- [german-ocr-3.1](https://huggingface.co/keyvan-ai/german-ocr-3.1) (2B) · [german-ocr-3](https://huggingface.co/keyvan-ai/german-ocr-3) (0.8B) – image-text-to-text models for German documents
- [german-ocr](https://huggingface.co/keyvan-ai/german-ocr) · [german-ocr-2b-gguf](https://huggingface.co/keyvan-ai/german-ocr-2b-gguf) · [german-ocr-turbo](https://huggingface.co/keyvan-ai/german-ocr-turbo) · [g-ocr](https://huggingface.co/keyvan-ai/g-ocr) – earlier generations and GGUF builds
**Vision and industrial inspection**
- [belegant-4b](https://huggingface.co/keyvan-ai/belegant-4b) – 4B image-text-to-text model for receipts and business documents
- [AnomalyDetection-MVTech-Metal](https://huggingface.co/keyvan-ai/AnomalyDetection-MVTech-Metal) – image segmentation for surface anomalies on metal parts
**Audio**
- [wakecore](https://huggingface.co/keyvan-ai/wakecore) – wake-word and audio classification
**Security: intrusion detection for vehicle and IT networks**
- [CanDefender](https://huggingface.co/keyvan-ai/CanDefender) · [CanDefender-Fuzzy](https://huggingface.co/keyvan-ai/CanDefender-Fuzzy) · [CANDefender-DoS](https://huggingface.co/keyvan-ai/CANDefender-DoS) – CAN-bus intrusion detection (fuzzing and DoS attacks)
- [SecIDS-v2](https://huggingface.co/keyvan-ai/SecIDS-v2) · [SecIDS-CNN](https://huggingface.co/keyvan-ai/SecIDS-CNN) – network intrusion detection
## Principles
Every published result comes with the protocol it was measured under, the controls that were run, and the limitations we found. Models are released under their base-model licences; reports under CC BY 4.0.
[keyvan.ai](https://keyvan.ai) · [GitHub](https://github.com/keyvanhardani) · [LinkedIn](https://www.linkedin.com/in/keyvanhardani)