Model Navigator
My current models: ASR fine-tunes, small classifiers and specialist models.
Daniel Rosehill
I'm based in Jerusalem, Israel, and I build AI systems, mostly around workflow automation, agent orchestration and speech. This is my home page on Hugging Face: my work here is spread across my own account and a few organisations, so these cards bring it together.
Browse my work
Dataset Navigator
All my datasets: speech, training data, Israel, images, reference and prompts, with audio and sentence navigators inside.
My Benchmarks
Benchmarks and evals, with the datasets, Spaces and GitHub repo of each project.
Everything on Hugging Face
The full list of models, datasets, Spaces and collections, on my AI site.
Collections
Every public collection, with a page for each.
Businesses and projects
Business
DSR Holdings
My consultancy: technical writing, communications and general consulting.
Business
Carrot Cake AI
AI consulting and agentic AI. A DSR Holdings company.
Project
Hebrish
On-device English-Hebrew code-switching dictation for Android, Linux and Windows. Hebrish on Hugging Face.
Project
My Weird Prompts
An AI-human podcasting experiment. My Weird Prompts on Hugging Face.
What I'm interested in
Speech recognition
Fine-tuning ASR for real dictation: technical vocabulary, English with Hebrew words in it, and models that run on the device.
Agentic AI and MCP
Agent orchestration, workflow automation and the tooling that makes agents practical to use.
Small task models
Classifiers and small models that do one job well and run locally, such as paragraph breaks or spotting a new task in an agent session.
Evaluation
Testing models for a real job: speech accuracy, image generation, local inference and how models behave.
Context and memory
Personal context for AI that stays under the user's control.
Human-AI collaboration
Open experiments in working and creating with AI, from podcasting to shared tooling.