How Mira Works
Mira uses semantic matching to understand implicit hiring criteria. Its agent autonomously searches, evaluates, and shortlists candidates from multiple sources.
Mira handles execution so you can focus on decisions. You describe what you need, the Agent does the searching and matching, and you review the results. Here's how the sourcing workflow works step by step.
You describe a role
↓
Agent clarifies requirements (only when details are missing)
↓
Agent searches across multiple talent sources
↓
MRE evaluates and ranks candidates
↓
Shortlist delivered with AI summaries
↓
You review, refine, or continueStep 1: You describe the need
Type a job description, paste a JD, or describe the role in natural language. Mira understands implicit criteria, you don't need formal search syntax.
Step 2: Agent clarifies (only when needed)
If your description is already specific, the Agent skips ahead. If key details are missing or ambiguous, it generates a short clarification form covering seniority level, number of candidates, and specific preferences. The form only appears when it would improve results.
Step 3: Multi-source search
Mira searches across three layers of data:
| Layer | Source | Description |
|---|---|---|
| Public data | Aggregated from multiple external talent databases | Broad coverage across industries and regions |
| Mira's own data | Mira's own index of public professional profiles and job postings | Continuously updated, globally scoped |
| Private data | Your uploaded resumes or ATS-integrated candidate pools | Your proprietary talent database, searchable by Mira |
Step 4: Reasoning-based matching
Mira Reasoning Embedding evaluates each candidate using multi-hop reasoning, not keyword matching:
- "At least 5 years of experience", calculates from career timeline.
- "From 0 to 1 experience", interprets as early-stage company building.
- "Public company background", infers organizational context from company data.
Step 5: Shortlist delivery
You get a curated list with:
- Candidate name, current role, company, and location.
- Years of experience.
- AI-generated summary of qualifications and fit.
- Full work history and education.
- Contact info, LinkedIn (always), plus email and phone where data exists.
- Downloadable data file.
Step 6: Iterate
Continue the conversation:
- Ask for more candidates.
- Adjust criteria ("also include Amsterdam", "prefer startup backgrounds").
- Start a new search for a different role.
Each task is saved in the sidebar and can be revisited anytime.
What's next
- Key Concepts: Agent, Shortlist, and other terms explained.
- Quick Start: Run your first sourcing task.