Candidate Evaluation Reports
Every candidate who passes through DeepScreen's funnel generates a structured evaluation report. These aren't just score summaries — they're comprehensive evidence packages that combine semantic resume analysis, proctored coding performance, AI interview transcripts, and integrity verification into a single actionable document for hiring managers.
Report Components
Resume Summary
Automated extraction of education, experience timeline, key skills, and project highlights. Semantic embedding captures contextual relevance beyond keyword matching — distinguishing a candidate who 'used React' from one who 'architected a React design system adopted by 40 engineers'.
Coding Assessment
Judge0 CE execution with OS-level proctoring. Scores include correctness, time complexity analysis, and code quality metrics. Proctoring confidence score flags AI copilot usage, tab switches, and suspicious process activity during the assessment window.
AI Interview Transcript
Full transcript of the AI avatar interview stage. Adaptive questioning paths logged with LLM reasoning traces. Responses scored against role-specific competency rubrics with confidence intervals rather than single-point estimates.
Proctoring Evidence Pack
Timestamped evidence from Tauri OS-level monitoring: gaze tracking logs, head pose deviations, detected objects (phone, tablet, second person), and blocked application attempts. Recruiter can review flagged segments without watching the full session.
Composite Score
Weighted fusion of all assessment signals into a single ranked score. Weights are configurable per role — technical roles weight coding higher, client-facing roles weight interview communication higher. Normalised against the candidate pool for relative ranking.
Scoring Methodology
| Component | Weight | Method |
|---|---|---|
Resume Fit | 25% | Section-wise cosine similarity against job description embeddings. Four weighted fields: Experience (0.35), Skills (0.30), Projects (0.20), Summary (0.15). |
Coding Score | 30% | Judge0 test case pass rate × code quality modifier. Penalty multiplier for proctoring flags. Threshold gate before advancing to interview stage. |
Interview Performance | 30% | LLM-evaluated responses scored on technical depth, communication clarity, and problem-solving approach. Adaptive difficulty ensures comparable scores across candidate pools. |
Integrity Score | 15% | Fused from gaze tracking, head pose, object detection, and process monitoring. Lower integrity scores reduce composite ranking regardless of other metrics. |
Export Formats
Per-candidate PDF with full assessment breakdown. Designed for hiring manager review with executive summary, detailed scores, and evidence links. Exportable as PDF or JSON for ATS ingestion.
Side-by-side comparison of top-K candidates for a single role. Radar charts show relative strengths across resume fit, coding ability, interview performance, and integrity score. Sorting by any dimension.
Aggregated weekly report showing all candidates assessed, conversion rates between stages, average scores by role type, and flagged candidates requiring immediate attention.