Pipeline builder
Compose resume, questionnaire, assessment and other module types into one ordered pipeline. Pin a version so in-flight candidates are unaffected by edits.
AI interview platform for technical hiring
Hire for evidence, not impressions.
Build a pipeline from resume, questionnaire and assessment rounds, invite candidates, and review a ranked shortlist where every score points back to the resume line, answer, or event that earned it.
- Every score cites evidence
- No candidate is auto-rejected
- Consent before any capture
Ranked shortlist
Sample product dataSenior Backend Engineer (Node/TypeScript)
Ranked by role-weighted score
| Rank | Candidate | Score | Dimensions | Status |
|---|---|---|---|---|
| 1 | Priya Sharmaincludes stub data role_fit89 relevance72 communication83 technical_depth69 | 76.5 | role_fit89 relevance72 communication83 technical_depth69 | done |
| 2 | Marcus Webbincludes stub data role_fit67 relevance66 communication71 technical_depth62 | 65.8 | role_fit67 relevance66 communication71 technical_depth62 | done |
| 3 | Dana Kowalskiincludes stub data role_fit50 relevance73 communication53 technical_depth66 | 62.0 | role_fit50 relevance73 communication53 technical_depth66 | done |
| 4 | Felix Aubertincludes stub data role_fit66 relevance47 communication59 technical_depth49 | 53.7 | role_fit66 relevance47 communication59 technical_depth49 | done |
Medium confidenceNeeds human review
Strong backend signal in the resume, with clear ownership of production systems. Communication evidence is thinner, so the score is marked for review.
Evidence
- resume_quoteStaff Backend Engineer, Fintail (2021–present)
- resume_quoteLed the migration of the payments service to Node/TypeScript
resume.score_vs_jd@v1 · model mock
Every row expands to its dimension scores, evidence and provenance.
SQLAssessment · SQL
28:14Return each country's total order value for 2025, highest first.
Your querySaved 14:32
SELECT c.country,
SUM(o.total_cents) / 100.0 AS revenue
FROM orders o
JOIN customers c ON c.id = o.customer_id
WHERE o.created_at >= '2025-01-01'
AND o.created_at < '2026-01-01'
GROUP BY c.country
ORDER BY revenue DESC;Run
Result · 4 rowsread-only
| country | revenue |
|---|---|
| IN | 184,320.00 |
| US | 92,110.50 |
Assessment · MCQAssessment · Question 2
41:08Round progress2 of 4
Which clause filters rows before grouping?
Pick the clause that filters individual rows before an aggregate is calculated.
- WHERE
- HAVING
- ORDER BY
Autosave is onSubmit round
How it works
See the full flowOne pipeline, from invite to evidence.
Compose the rounds that predict performance for the role, invite candidates with one link, and review a shortlist built from scores you can inspect.
- 01
Compose the pipeline
Pick the modules that fit the role and put them in order. Versioning keeps a candidate on the pipeline they started.
- 02
Invite with one link
Send an individual single-use link or a public link with an expiry and a use cap.
- 03
Candidates sign in and take rounds
One portal, one clear next step, a visible deadline and autosave throughout.
- 04
Grade with evidence
Deterministic parts stay deterministic. LLM grading returns structured scores, reasoning, confidence and evidence.
- 05
Review the ranked shortlist
Open a candidate to see the dimension breakdown, evidence, confidence and any override history.
The platform
Explore all featuresThe screen, the evidence, and the candidate experience in one place.
boson keeps the recruiter workspace and the candidate portal on the same pipeline, so nothing gets lost between the apply button and the shortlist.
Ranked shortlist
See candidates ordered by a per-job score with the dimension breakdown behind it, including how much of the result comes from stub or low-confidence rounds.
Evidence-backed scorecards
Each dimension carries reasoning, evidence, confidence and provenance. Open a score to see the quote, answer or event that produced it.
Resume review with receipts
A resume is graded against the job description across relevance, technical depth, communication and role fit. Quotes are verified against the source text before they are shown.
Questionnaire
Written questions in the candidate portal. Choice answers and required-field checks are deterministic; free text is graded with structured LLM output, evidence and confidence.
MCQ and SQL assessment
Multiple-choice and SQL questions with deterministic scoring. The candidate SQL preview runs read-only against a throwaway in-memory database.
Explainable grading
A score is only useful if you can see why.
Each dimension carries a score, reasoning, confidence, and evidence you can inspect. When the model is unsure, the run is flagged for human review instead of being silently rejected.
Verbatim evidence
Resume quotes are checked against the source text. Unverified quotes are dropped and logged.
Confidence, not certainty
Multi-sample judging reports a median and spread per dimension, so thin evidence is visible.
Provenance on every score
The prompt version and model version are stored with the score, alongside any human override.
Humans make the decision
Steps can be required, but outcomes are scores and flags. Nobody is rejected automatically.
Resume, questionnaire and assessment rounds run end to end today. Coding, system design, voice and human interview rounds share the same scoring backbone as they ship.
ScorecardNeeds human review
Resume review
Senior Backend Engineer (Node/TypeScript)
Overall score
76.5
role_fit89
relevance72
communication83
technical_depth69
Evidence
- resume_quoteStaff Backend Engineer, Fintail (2021–present)
- resume_quoteLed the migration of the payments service to Node/TypeScript
resume.score_vs_jd@v1 · model mock
Sample data. In the product, every score links to its evidence and provenance.
Trust and privacy
Consent first. Evidence always. Humans in control.
Technical hiring touches people's livelihoods. The product is built so candidates know what is being evaluated, and recruiters can defend every decision.
Available today
Consent gate
Candidates consent to AI evaluation before a run starts. Recording consent is separate per capture type.
Available today
Text, not photos
Grading uses structured answers and text. Photos are not used, and questionnaire grading strips names, contact details and protected attributes from résumé context.
Available today
Human in the loop
Overrides require a reason and leave an audit trail. Low-confidence scores are flagged for review.
Available today
Access and retention
Recordings carry a retention date, access is logged, and candidates can submit a deletion request from the portal.
Coming soon
Fairness reporting
Score distribution and drift reports where lawful group data is available. This is coming soon, not shipped today.
Available today
Candidate visibility
Candidates always see their status and next step. Result visibility is a per-job recruiter setting, off by default.
Pricing
View pricingPilot pricing is being shaped with the first teams.
boson is in active development. Packaging is intentionally not final, so the pricing page shows illustrative tiers and marks every number as a placeholder.
Pilot
Illustrative tierTBD
pricing model not decided
Team
Illustrative tierTBD
pricing model not decided
Scale
Illustrative tierTBD
pricing model not decided
No plan is on sale yet. The pricing model is undecided and will be scoped with pilot teams.
Contact optionsFAQ
More questions? Contact usQuestions worth asking before you trust a score.
Straight answers about consent, evidence, AI use, and what is actually shipping today.
Start with one role
Run a screen you can explain.
Build a pipeline, invite candidates, and review a ranked shortlist where every score has a receipt. Start with resume, questionnaire or assessment rounds, and add deeper modules as they ship.