Search Filters & Criteria
Understand the difference between hard filters and AI-powered criteria in Mira.
When you describe a role to Mira, your requirements are processed in two ways: as hard filters and as AI criteria. Understanding the difference helps you write better descriptions and get more accurate Shortlists.
Hard filters vs. AI criteria
| Hard filters | AI criteria | |
|---|---|---|
| What they do | Strictly include or exclude candidates | Influence ranking and matching analysis |
| Examples | Location, language, minimum years of experience | "Startup DNA", "has built from 0 to 1", company type preference |
| How they work | Binary, candidate either passes or doesn't | Gradient, candidates are ranked by degree of match |
| When to use | Non-negotiable requirements | Strong preferences that allow flexibility |
Hard filters
These are requirements that candidates must meet. If a candidate doesn't match a hard filter, they won't appear in your Shortlist:
- Location: "Must be in Berlin", only Berlin-based candidates.
- Language: "Must speak German", only German-speaking candidates.
- Minimum experience: "At least 5 years", excludes anyone with less.
AI criteria
These are preferences that Mira Reasoning Embedding evaluates through reasoning. Candidates who match these criteria rank higher, but candidates who partially match may still appear:
- Company type: "Preferably from a high-growth startup", startup backgrounds rank higher, but strong candidates from other environments still surface.
- Implicit qualifications: "Has built a team from scratch", The MRE looks for leadership patterns across career history.
- Cultural signals: "Scrappy, hands-on", The MRE identifies candidates from environments that typically develop these qualities.
How the clarification form uses both
When the Agent generates its clarification form after your initial input, it's essentially separating your requirements into hard filters and AI criteria:
- Seniority level → Sets a hard filter or strong AI criterion depending on your choice.
- Number of candidates → Controls search scope.
- Company type preferences → Sets AI criteria for ranking.
Tips
- Start with fewer hard filters. You can always narrow down after seeing the first Shortlist. Over-filtering upfront may exclude strong candidates.
- Use AI criteria for "nice-to-haves." Instead of "must be from a startup," try "preferably from a startup", you might discover great candidates from unexpected backgrounds.
- Refine in conversation. After seeing results, tell the Agent: "Only show me candidates from startups" to tighten criteria, or "Broaden to include enterprise backgrounds" to relax them.