inVision U Admissions · Powered by AI

We don't find
the best writers.
We find the next leaders.

A multimodal admissions pipeline — video, personality test, and AI-powered analysis — with full explainability and human reviewers making every final call.

Video Presentation3-min recorded intro
Personality Test40 leadership questions
AI AnalysisExplainable scoring
Human DecisionReviewer makes final call
0Scoring Pillars
0Bias Tests
0%Explainable
ZEROAuto-rejections
Process

The Admissions Flow

Five steps. Multimodal signals. One human decision.

1

Apply

Choose Foundation Year or Undergraduate and fill in your profile.

2

Video Presentation

Record a 3-minute video in your own words — authenticity over polish.

3

Personality Test

40 leadership questions revealing how you think, adapt, and respond.

4

AI Analysis

Scores across 4 dimensions with full evidence tied to your responses.

5

Human Review

Admissions reviewers see all evidence and make the final decision.

Scoring

Four Pillars of Potential

Weights derived from 17 meta-analyses. Every score backed by evidence quotes from the candidate's own words.

Leadership Potential

Video presentation

30%

Ability to mobilize others, take initiative, and drive collective outcomes. Weighted highest because inVision U seeks future leaders — but reduced from 35% due to higher measurement noise in essay/video contexts.

Moderate but noisySedlacek NCQ (2004); Hogan & Kaiser 2005; ETS PPI

"I walked through how I coordinated 40 volunteers after the flood — logistics, morale, and follow-through."

Growth Trajectory

Personality test + application

25%

Rate of improvement and learning velocity, not just current achievement. Maps to conscientiousness — the most robust personality predictor of academic success (r = 0.21).

r = 0.19–0.21Crede et al. 2017; Poropat 2009; Mammadov 2022

"Strong self-awareness of past failures; a jump from failing math to tutoring peers within two semesters."

Authentic Motivation

Bot interview (STAR)

28%

Genuine intrinsic drive to create impact beyond personal gain. Has the strongest evidence base of any non-cognitive predictor — identified regulation is the best predictor of persistence.

ρ = 0.30Robbins et al. 2004; Howard et al. 2021; Cerasoli 2014

"When asked what they'd do with no audience, they described teaching coding to younger students — unprompted and specific."

Resilience & Agency

Video presentation

17%

Capacity to overcome adversity and proactively shape one's path. Critical for surfacing "hidden talent" — strong candidates who self-present poorly but overcome systemic barriers.

r = 0.16Resilience meta-analyses 2023; Sedlacek 2004

"After my family lost everything, I didn't wait — I built a small resale business at 16 to cover my own school supplies."

Formula

How the Composite Score Works

A weighted average — not a black box. Every dimension contributes proportionally.

Composite=Leadership×30%+Growth×25%+Authentic×28%+Resilience×17%

≥ 4.0

Shortlist

AI recommends → human confirms

2.5 – 3.9

Needs Review

Mandatory human evaluation

< 2.5

Flagged

Human can still override to admit

Scale: 1.0 (Absent) to 5.0 (Exceptional) per dimension. No candidate is ever auto-rejected — humans make every final call.

Personality Test

7 Dimensions → 4 Pillars

The personality test measures 7 raw dimensions that are combined into the 4 composite pillars above.

Leadership+InitiativeLeadership Potential

max(leadership, initiative)

Growth Mindset+Critical ThinkingGrowth Trajectory

avg(growth_mindset, critical_thinking)

Integrity+EmpathyAuthentic Motivation

weighted(integrity × 0.6, empathy × 0.4)

ResilienceResilience & Agency

avg(resilience)

Evidence

Research-Backed Scoring

Every weight in our model is grounded in peer-reviewed meta-analyses. Click any card to read the original paper.

7Peer-reviewed papers
700K+Total participants studied
750+Independent samples
5Top-tier journals
All citations verified via PubMed & ERIC — links go directly to source

How This Translates to Our Weights

28%

Motivation

Strongest non-cognitive predictor of GPA (ρ = 0.30) and best predictor of who stays enrolled. Intrinsic drive matters more than test scores.

30%

Leadership

Highest because inVision U seeks leaders — but limited to 30% because leadership measurement from video/transcript is noisier than other dimensions.

25%

Growth

Maps to conscientiousness (r = 0.19-0.21), the most stable personality predictor of academic performance across 413,000+ participants.

17%

Resilience

Smaller effect size (r = 0.16) but critical for finding hidden talent — strong candidates from tough backgrounds who don't self-present well.

Weights are expert-derived, not regression-optimized, to avoid overfitting on small samples. The compensatory model (strengths offset weaknesses) was validated by Ock & Oswald (2018) as producing higher criterion performance than threshold-based approaches.

Ethics

Built for Fairness, Not Just Speed

Transparency and human oversight at every stage of the pipeline.

Human-in-the-loop for every decision

AI recommends, humans decide. No candidate is ever auto-rejected without expert review.

5 computational bias tests

Name-blind, region-blind, gender-neutral, distribution, and correlation audits run on every batch.

Full explainability across all signals

Every score is backed by direct evidence — video quotes, personality responses, and application text.

Multimodal, not text-dependent

Strong writers don't automatically win. Video, structured tests, and adaptive interviews are weighted equally.

AI does NOT make final decisions
Architecture

Built With

Frontend

  • Next.js 15
  • React 19
  • TypeScript
  • Tailwind CSS v4
  • shadcn/ui
  • Motion

Backend

  • Python 3.12
  • FastAPI
  • LangGraph
  • aiogram 3.x

AI / Cloud

  • Azure OpenAI (GPT-5.4-mini)
  • Azure Speech STT
  • Azure Language
  • Azure Document Intelligence

Data

  • PostgreSQL + pgvector
  • JSON Cache
  • Embeddings (text-embedding-3-small)