PROJECT — 2026

An AI college-admissions planning platform — committee-grade application review without the private-consultant price tag.

Trajecta is a full-stack admissions planning platform I co-founded and lead engineering for: a React/Vite front end over an Express/TypeScript API, Prisma on the data layer, Supabase handling auth. It exists because committee-quality application feedback is mostly locked behind private consultants.

The core is a multi-agent LLM committee: several simulated admissions readers evaluate an application in parallel, constrained by fixed JSON output schemas and server-side validation — the models argue inside guardrails rather than free-associating. A six-category profile-scoring and college-chancing engine calibrates their outputs against Common Data Set percentiles, correcting the self-report bias that plagues chancing tools.

Documents enter through a PDF import pipeline that strips metadata and enforces schema constraints before anything reaches a model: untrusted uploads are treated as a prompt-injection surface, not just files.

React · TypeScript · Express · Prisma · Anthropic API