Six surfaces. One pipeline.
A working capability showcase, built in one week. 50 students, 200 counselling reports, 250 career matches — pulled from a live Postgres backend, searched, ranked, and explained by AI.
Capabilities
— 6 surfacesFind any student. Postgres FTS over 200 counselling reports, filtered by school, grade, RIASEC, kind.
Cohort metrics. Aptitude distributions, RIASEC profiles, career-fit aggregates, school comparisons.
Per-student AI summaries with citations. Multi-modal — narrative + RIASEC radar + statistical charts.
CSV / PDF / image upload. LLM auto-tags content, extracts structured fields, flags anomalies.
ICS calendar feed + Whisper-transcribed sessions + AI-generated counselling notes.
Role-based filtering (counselor / admin / parent / reviewer) with RLS-style query gates and an audit log.
Most demos skip §11. We demo all three.
The RFP closes with: “the demo should include handling of edge cases such as incomplete data, access restriction conflicts, and updates to underlying data.”
- Incomplete data — Sneha (#50) is missing 3 RIASEC dims + 2 aptitudes. Summary still generates with a precise “what's missing” note + MEDIUM confidence.
- Access conflict — viewer tries to delete a session → 403 page renders the actual request + the audit_log row that was just written.
- Data update — admin edits a student name → /discovery shows refresh banner; /insights flags the now-stale summary.
From raw data to actionable career direction
Upload CSV exports, PDF reports, scanned notes, audio sessions. LLM auto-extracts structured data.
Auto-tag, score, classify. RIASEC matching, career-fit calculation, FTS index build, demand-weighted ranking.
Search across the corpus. Filter by school, grade, RIASEC, kind. Counselors find evidence in seconds, not minutes.
Per-student AI summary with citations. Counsellor receives evidence-backed recommendations they can stand behind.