An interview is more than a conversation. Abilint gives your candidate real tasks in a real cloud — then rates what they actually did, on evidence both sides can see.
ability, linted.
$ abilint assess --candidate jane --cloud aws --cap $15 ✓ path: 4 depth-problems (CV-driven) ✓ real cloud: aws · workstation + budget $15 · ttl 45m ✓ vera linting the evidence (git + checks + timeline)… security_iam L3 ████████░░ 85 iac_terraform L2 ██████░░░░ 65 networking L0 ░░░░░░░░░░ · … ✓ rating: 75/100 · seniority: Mid · integrity: high → evidence pack + rating delivered to both sides
Algorithm puzzles and simulated IDEs can't show who can actually run production. We ask a candidate to fix something and build something — in their own real cloud account, on problems scoped to what their resume claims. You pick what matters; they prove it.
Vera reads the CV and builds a transparent path of depth-problems, rising to the claimed stack.
An ephemeral account is vended (AWS member / Azure sub / GCP project). The interviewer sets the cost cap.
Code diffs, infra snapshots, check results + an action timeline — an immutable, hash-stamped bundle both sides receive.
A trustworthy profile: per-axis depth, percentile, seniority, and an anti-gaming integrity signal.
It's up to the reviewer and their budget. Both produce the same Vera Rating + evidence pack.
Real-time pair-session in real cloud. Quick problem-solving under time pressure. Best for first-pass screening.
The candidate receives a brief — real tasks in real cloud, a budget, a deadline — and works autonomously. Then comes a walkthrough: they present their work, you probe depth. Promotes discussion.
Not a single vibe-score. A profile across the role's competencies, with the evidence behind every tier — derivable, auditable, comparable.
AWS, Azure, GCP — the candidate's own native account, not a simulator. Each session isolated + cleanly billed.
Problems rise to what their CV claims; the interviewer can't dodge the must-probe skills.
A hash-stamped bundle the employer and the candidate receive — nothing happens in a black box.
Cloud at cost + 100% (it's our risk to clean up). You set the cap; the session auto-stops at TTL or spend.
The pair-session isn't surveillance — it's collaboration, in the live environment.
Vera checks the work for paste-detection + cadence — so the rating reflects real work, not copied code.
$50 flat per candidate assessment (the platform + rating fee). Plus the candidate's actual cloud cost × 2 — the markup covers our cleanup, orphan liability, and breakage risk. You set the cloud cap per candidate; the total is always $50 + (cloud ≤ cap) × 2.