Direct answer
AI/ML sourcing works best when you search for evidence: frameworks, model work, papers, repos, evaluations, deployment context, and productized systems.
Signals that matter
Look for PyTorch, TensorFlow, transformers, embeddings, RAG, vector databases, model serving, evaluation, fine-tuning, CUDA, Triton, and production ML infrastructure.
Operating notes
- Search frameworks plus deployment context.
- Check recency and ownership of repos.
- Pair research evidence with engineering evidence.
- Verify production experience directly.
Where to search
Use GitHub for code, Hugging Face for model and dataset evidence, OpenAlex for research, personal sites for project context, and LinkedIn for employment history.
False positives
Many profiles mention AI without production depth. Separate prompt-use, coursework, and tutorials from shipped ML systems.
SourcingOS workflow
Candidate Search can route AI/ML queries toward GitHub, Hugging Face, OpenAlex, npm, and PyPI lanes.
Copy-paste starting strings
(PyTorch OR TensorFlow OR JAX) AND ("model serving" OR Triton OR Kubernetes OR "production ML")site:huggingface.co (transformers OR embeddings OR "text generation")
site:github.com (RAG OR embeddings OR "vector database") (Python OR TypeScript)
FAQ
Should I search for AI Engineer?
Yes, but pair it with specific evidence terms. Titles alone are noisy.
Is a paper enough evidence?
It can be strong research evidence, but confirm engineering and product context for production roles.