Jason Liu
Instructor — typed, validated structured output from language models.
2 evidence links
← Discover
Prompting
Hire for prompt design that can be tested
Instruction design, structured outputs, and evals — the work that makes a model follow a job description.
4Published profiles in this specialty
100%Of those rows attach evidence
8Capabilities in the taxonomy
Why hire through this index
Prompting on this directory is a capability, not a personality. Profiles that list it are expected to show how they specify a task and how they know it still works after a model change.
Available prompting
4 published profiles listing this capability.
Omar Khattab
ColBERT and DSPy — late-interaction retrieval and compiled LLM pipelines.
2 evidence links
Simon Willison
Datasette and the llm CLI, plus a public archive of prompt-injection notes.
2 evidence links
Mina ExampleDemo
Prompting and guardrails for tasks that must not hallucinate.
Toronto · Selectively available · 2 evidence links
Frequently asked
Isn’t prompt engineering just writing good instructions?
Not at production scale. The work is version control, eval harnesses, regression tests, and observability. That is why prompting sits next to evaluation and guardrails in this taxonomy.
Can they help with safety and guardrails?
Many prompting profiles also list guardrails — jailbreak resistance, output validation, red-teaming. Use the capability filters on Discover to require both.
Do they work across model families?
Strong prompt work is model-agnostic: the same eval suite against more than one host. Read the evidence; do not take a single-vendor claim at face value.