Will our students' decks train your model? Who sees them?

Never trained on — contractual. Submissions are processed only for your competition, in a closed perimeter with no public links; the sub-processor and model-provider list is available for your IT review, and the institution owns every report and score.

Won't teams say “an AI judged us”?

State it before they ask — that's what the disclosure kit is for. The honest framing flips the optics: every entry got a full read under identical rules, which is more than any volunteer process can claim, and humans made every ranking decision. What's genuinely hard to defend is entry #300 getting a tired judge at 11pm.

Judging is how we engage sponsors and alumni.

Keep every judge — change the ask. Nobody sponsors your event to do six hours of reading homework; they come to see finalists and be seen. Remote screeners keep their touchpoint as a one-hour confirm-and-comment pass, and the same judges return next year.

We run multiple tracks with different rubrics.

Tracks, criteria and weights are configured per competition; each track scores against its own rubric, and the leaderboard respects your weighting.

Can a team appeal or see their own record?

Two things are product defaults: a team can request its own record, and no entry is eliminated without a named human sign-off. Whether you run open appeals stays your policy — the per-entry evidence, scores and judge decisions make both a five-minute conversation instead of an archaeology dig.

How do we validate this before it touches a real result?

The shadow pilot: your screeners run one written round as usual while the AI panel reads the same pool. You compare coverage, hours and where the pre-ranking agreed with your screeners — validated against your own judges before it influences anything.

Will our students' decks train your model? Who sees them?

Never trained on — contractual. Submissions are processed only for your competition, in a closed perimeter with no public links; the sub-processor and model-provider list is available for your IT review, and the institution owns every report and score.

Won't teams say “an AI judged us”?

State it before they ask — that's what the disclosure kit is for. The honest framing flips the optics: every entry got a full read under identical rules, which is more than any volunteer process can claim, and humans made every ranking decision. What's genuinely hard to defend is entry #300 getting a tired judge at 11pm.

Judging is how we engage sponsors and alumni.

Keep every judge — change the ask. Nobody sponsors your event to do six hours of reading homework; they come to see finalists and be seen. Remote screeners keep their touchpoint as a one-hour confirm-and-comment pass, and the same judges return next year.

We run multiple tracks with different rubrics.

Tracks, criteria and weights are configured per competition; each track scores against its own rubric, and the leaderboard respects your weighting.

Can a team appeal or see their own record?

Two things are product defaults: a team can request its own record, and no entry is eliminated without a named human sign-off. Whether you run open appeals stays your policy — the per-entry evidence, scores and judge decisions make both a five-minute conversation instead of an archaeology dig.

How do we validate this before it touches a real result?

The shadow pilot: your screeners run one written round as usual while the AI panel reads the same pool. You compare coverage, hours and where the pre-ranking agreed with your screeners — validated against your own judges before it influences anything.

AI prepares the analysis · People decide

Every entry gets a full read. Every rank carries its receipts.

EvalLens is the judging layer under your existing competition: an AI panel pre-reads the whole written round on your rubric, your screening committee confirms with briefing packs instead of blank PDF stacks, and your live judges walk into semis briefed. The final call stays exactly where it belongs.

Run a shadow pilot See a sample report

The written round, honestly

Three battles every director fights

300+volunteer judges one flagship university competition publicly recruits — every single yearrbpc.rice.edu
4–6 hthe real homework in one screening assignment: three written entries with feedback write-upswhy good judges decline next year
~5%share of the pool the average judge actually sees at a large event — scores barely compareHackMIT judging research
0teams who see the written-round feedback at most major competitions todayit goes to the director's drawer

The Monday-after email

“Judging process concerns” — cc: sponsor, cc: dean.

A faculty advisor’s team lost. Today you answer with a shrug and an apology draft, because the written round genuinely is a lottery of tired volunteers. With a record, the thread dies in one reply.

Score

7.4

on “business model” — your rubric, your weights

Finding

Pricing validated with early customers; distribution still a hypothesis.

Quote

“…14 pilot customers at $190/mo…” · page 9

Evidence strength

Strong — 3 of 3 claims verified on-page. No quote, no finding.

Judge scores

Human panel's scores and deliberation notes, logged alongside.

The first and the last entry are evaluated under exactly the same rules — something no volunteer process can honestly claim. Judge disagreement surfaces for deliberation — never averaged away.

How it works

Applications open → awards night, in six steps

Prelim decisions stay human. The AI panel does the first read; your judges keep the room and the final ranking.

01
Your rubric and tracks, locked
Criteria, weights and tracks configured per competition — plus a methodology statement you can publish in the competition rules, where fairness claims legally live. You get: a rulebook and a public fairness statement.
02
Entries land on your page — or alongside your platform
Link or QR, open or invite-only, with a window and live statuses. Completeness is checked automatically; staff chases exceptions only. You get: a clean pool before judging starts.
03
The AI panel does the first read of every entry
Six reviewer roles — independent AI reads, not people — score the whole pool on your rubric, every page, coverage logged. One panel for the whole pool: no judge-assignment lottery in the written round. You get: the written round pre-read in hours.
04
Your screeners confirm — with packs, not piles
Prelim decisions stay human. Your remote judges and alumni keep their role: screening becomes a one-hour confirm-and-comment pass instead of a six-hour reading weekend. Same touchpoint, a fraction of the ask — and they say yes again next year. You get: a judge pool that renews itself.
05
Live rounds run exactly as designed
Semis and finals: same stage, same judges, same drama. Judges score as usual; the leaderboard is built from human scores times your weights; disagreements surface for deliberation. You get: your judges' leaderboard, better informed.
06
Awards night — and every team leaves with something
Per-rank record for anyone who asks; structured feedback for all teams, reviewed by your staff before it's sent. For a student competition, feedback for 400 teams is a teaching outcome, not just an event. You get: teams that return and recommend.

Judge math

1,800 hours, returned to the parts judges love.

400 entries × 3 reads × 1.5 hours = 1,800 judge-hours your current written round consumes. The AI panel returns those hours to semis, finals and mentoring — and every entry still gets every page read.

Keep every judge

Judge count is a program KPI and a sponsor perk. Nothing here reduces it — it upgrades what the ask is.

One standard, whole pool

The written round stops being a lottery of which tired judge drew your entry on which night.

Feedback as pedagogy

Every student team gets rubric-based feedback — the outcome deans and sponsors actually brag about.

When the student paper calls

You get the script, not just the software.

The “what we tell everyone” kit, included.

The real fear isn’t the tool failing — it’s defending the tool with no script. So the setup includes ready-to-use language for every audience:

The on-stage sentence“Every entry received a full read under identical rules — and humans made every ranking decision.”

The rules-document paragraphA methodology statement for your competition rules: what AI assists, what humans decide, how teams can ask about their own record.

The team-facing disclosureA plain-language snippet for the application form, so no one discovers AI involvement after the fact.

The COI noteJudge conflict-of-interest and recusal handling stays your policy — the record simply logs who scored what, which makes recusals verifiable.

Data & student IP

The block your IT office reads first.

Never trained on

Student and team submissions are processed only for your competition's evaluation — never used to train models. Contractual.

The institution owns the record

Reports, scores and the decision log belong to your program. Retention and deletion on your policy; student-data handling structured to support your FERPA obligations, DPA available.

Procurement-ready

PO and invoice accepted · vendor registration forms and security questionnaires supported · public sub-processor list at /subprocessors · education discount for university programs.

Priced per event, not per seat — one event, a full season, or a small pitch night. PO and invoice accepted; education discount for university programs. See pricing or book a call.

FAQ

What your stakeholders will ask

Will our students' decks train your model? Who sees them?: Never trained on — contractual. Submissions are processed only for your competition, in a closed perimeter with no public links; the sub-processor and model-provider list is available for your IT review, and the institution owns every report and score.
Won't teams say “an AI judged us”?: State it before they ask — that's what the disclosure kit is for. The honest framing flips the optics: every entry got a full read under identical rules, which is more than any volunteer process can claim, and humans made every ranking decision. What's genuinely hard to defend is entry #300 getting a tired judge at 11pm.
Judging is how we engage sponsors and alumni.: Keep every judge — change the ask. Nobody sponsors your event to do six hours of reading homework; they come to see finalists and be seen. Remote screeners keep their touchpoint as a one-hour confirm-and-comment pass, and the same judges return next year.
We run multiple tracks with different rubrics.: Tracks, criteria and weights are configured per competition; each track scores against its own rubric, and the leaderboard respects your weighting.
Can a team appeal or see their own record?: Two things are product defaults: a team can request its own record, and no entry is eliminated without a named human sign-off. Whether you run open appeals stays your policy — the per-entry evidence, scores and judge decisions make both a five-minute conversation instead of an archaeology dig.
How do we validate this before it touches a real result?: The shadow pilot: your screeners run one written round as usual while the AI panel reads the same pool. You compare coverage, hours and where the pre-ranking agreed with your screeners — validated against your own judges before it influences anything.

Next step

Shadow-pilot your next screening round.

Applications flow as usual; the AI panel pre-reads in parallel with your screeners; you compare before anything counts. One event, fixed price, PO and invoice welcome. The first run is free through August 31, for batches up to 10 decks.

Set up the shadow pilot hello@evallens.io