Roughly 39 million Americans hold college credit and federal loan debt but no finished credential. Persist is the eligibility engine that tells each of them, and you, exactly which rule is blocking them and which pathway still works.
The stop-out problem
A near-completor is a student you already admitted, already taught, and already have a transcript for. They stopped one or two terms short, and now they carry the debt without the credential that would have paid for it. Re-enrolling one of them is cheaper than recruiting a new applicant, and it moves your completion rate instead of just your headcount.
What makes this urgent rather than perennial: the One Big Beautiful Bill Act (H.R.1) took effect July 1, 2026, and created a new set of stop-out triggers. Grad PLUS is gone for borrowers whose enrollment continuity is broken. Lifetime federal loan caps now apply to debt already borrowed, not just new borrowing. Both of those turn a returning student into an ineligible one, silently, and the student usually finds out from a denial letter rather than an advisor.
Which means a share of your own stopped-out population became harder to re-enroll this month, and nobody has the list.
What Persist is
Financial-aid eligibility is fragmented, effective-dated, and changes under people. So the core of Persist is a versioned encoding of the actual rules, and a deterministic evaluator that runs them. No language model sits in the decision path.
Each rule is a dated record with its statutory citation, so an answer given today can be reproduced and defended a year from now, and a rule change is a diff instead of a rewrite.
The same student profile always produces the same flags and the same pathways. Every output traces back to the specific rule that produced it. Nothing is generated or guessed.
Being blocked from federal borrowing is not the same as being unable to finish. The engine matches prior-learning credit, Workforce Pell, employer tuition benefits, and state completion programs against the student's actual situation.

See it run
grad_plus_elimination_2026lifetime_loan_caps_2026Every line she was told traces to a named rule. Here is the same profile run through the engine directly, which is what actually produced those four results.
Case for: Maria (composite, sample profile) (as of 2026-07-27) Flags: [BLOCKING] (grad_plus_elimination_2026) You have prior Grad PLUS borrowing, but a 431-day gap in enrollment breaks continuity under OBBBA, H.R.1, effective 2026-07-01. You're treated as a new borrower and Grad PLUS is not available to you, regardless of how close you are to finishing. [BLOCKING] (lifetime_loan_caps_2026) Total federal borrowing ($105,000) is at or over the new lifetime cap ($257,500 universal, $100,000 graduate-specific). This applies to debt already borrowed, not just new borrowing. [INFO] (pell_eligibility_cliffs_2026) SAI is under the $14,790 cutoff, Pell eligibility isn't blocked on this basis. Matched pathways: - [prior_learning_credit] ~3 weeks, ~$200, no new federal borrowing Petition the registrar for credit from professional experience or military training. State resources: - [NJ] Some College, No Degree (SCND) Initiative Ask the registrar or student success office whether they're an SCND grant recipient before assuming self-funded re-enrollment is the only option.
The pilot
No integration, no data feed, no procurement cycle. Start with a list your completion team already has.
Stopped out within the last 24 months, past a credit threshold you set. A few hundred students is plenty.
A short intake they fill in themselves. No FERPA-covered record leaves your systems for this stage.
Each student gets named blocking rules, viable pathways, and state programs. Your advisors get the same view, in one list.
Re-enrollment and pathway completion after one term, against the cohort you'd have written off. Then decide.
Why this fits NYU
Full-price graduate tuition leans on Grad PLUS. With it eliminated for continuity-broken borrowers, a returning graduate student can be academically welcome and financially blocked at the same time.
The new lifetime caps count debt already borrowed, so students who transferred in graduate credit from another institution can hit the ceiling on arrival. Combined-institution borrowing is exactly the case the engine checks.
SUNY/CUNY Reconnect, including the 2026/27 expansion to associate programs at eleven bachelor's-granting SUNY campuses, is in the state ruleset with its eligibility band and the question to ask.
Re-enrolling near-completors moves completion rates, which matters more under the program-level outcome accountability framework phasing in behind OBBBA. That rule is tracked in the engine as a watch item.
Data & compliance
The fastest way to make this project unworkable would be to start by asking for institutional data. So it doesn't.
Students enter their own enrollment history, borrowing, and state. No SIS, NSLDS, or transcript integration exists, which keeps the FERPA and GLBA surface near zero for a pilot.
Any future record-level feed happens only behind a signed school-official agreement and review by an education-data privacy attorney. That's a deliberate later decision, not something bundled into a pilot.
Every determination names the rule, its effective date, and its citation. If a rule changes, previously issued guidance can be re-run and diffed rather than trusted.
Output is written as "here is the rule that applies and the question to ask," never as an official eligibility ruling. Your aid office remains the decision-maker.
Let's talk
Twenty minutes to see whether one cohort and one term of measurement is worth running before the fall census.
Persist · an eligibility engine for near-completors · a separate project from Voyacover, same builder