Zobique: AI career-intelligence platform for universities

The constraint#

The gap between "graduated" and "employable" is an infrastructure problem, not a teaching one. A university has grades, attendance, and a placement cell that runs on spreadsheets. It does not have a per-student read on what specifically is between this person and the job they want, early enough to act on it.

Students discover the gap the hard way, through a run of rejected applications and failed interviews, at which point there is a semester or less to close it.

What I built#

Zobique is a multi-tenant platform that each university runs for its own cohort:

• Career Studio. A student's profile, target roles, and history feed a retrieval-augmented model that returns a concrete skill-gap read and a personalised path, ranked by impact on job readiness rather than a generic syllabus.
• Job & application intelligence. Matching against live postings, ATS-aware résumé feedback, and interview practice with structured feedback.
• Placement intelligence. The placement cell gets an aggregate view of where its cohort is weak, so intervention is a decision backed by data rather than a guess.

What broke#

• "Personalised" that wasn't. The first skill-gap output was too close to a generic roadmap. It only became useful once the ranking was tied to observed outcomes for similar profiles rather than to a fixed skill tree.
• Cold-start profiles. A first-year with almost no history gets a weak read. The honest fix was to say so in the UI rather than manufacture confidence.
• Placement-cell trust. Aggregate dashboards are only used if the numbers match what the cell already believes about its cohort on the cases they know well. Getting those anchor cases right was what earned the rest.

Outcome#

Piloted across 2 universities with 1,500+ students, contributing to 50+ placements.

What I'd redesign#