● AI Product Build — University Venture

Turning nine disconnected campus platforms into one AI orchestration layer.

Role — Technical Co-Founder & Backend Architect / Venture — Last Minute Club (AI Campus Orchestrator) / Timeframe — 2025–2026 · MSIN0230, UCL School of Management

Last Minute Club began as a team project for UCL's Artificial Intelligence for Business module: could an AI layer sit above the platforms students already use — Moodle, Portico, careers, the students' union, the library, their own calendar — and turn scattered inputs into a conflict-free week and an early warning on burnout? As technical co-founder, I designed the data model, the rule-based orchestration engine, the heuristic scoring layer, and the Claude-powered assistant on top of it, then validated the MVP against simulated schedules and a small live pilot.

Note on scope — a 3-person course venture with a working MVP and a small self-reported pilot, not a deployed university product. Framed throughout as early validation, not institutional-scale proof.
01 / The Problem

Orchestration is missing, not resources

Baseline, before any of this existed

University students juggle up to nine disconnected platforms — Moodle for assignments, Portico for grades, UCL Careers for jobs, the Students' Union for events, the library for research, a personal calendar for everything else. Each one stores data. None of them talk to each other, plan around a deadline, or notice a student heading for burnout before it happens.

28%
of students have considered dropping out
HEPI, 2024
41,914
student dropouts in 2022/23 — a record high
SLC, 2023
12%
feel their university handles mental health well
CALM, 2023
02 / Readings

The instrument panel

Benchmarked across five simulated student schedules and a small live pilot, 2025–2026.
0%
Scheduling Conflicts Reduced ▼
Rule-based orchestration engine replacing manual, ad hoc scheduling — vs. an unplanned baseline.
0%
Self-Reported Burnout Score ▼
Average drop across 4 pilot testers' before/after scores — a small self-reported sample, not a clinical result.
<0s
Avg. Schedule Generation Time
Full multi-week, conflict-free timetable produced end-to-end by the orchestration engine.
0
Burnout Risk Dimensions Modelled
Evidence-based risk factors assessed per generated schedule — heuristic, not clinical.
03 / Scope

Four connected workstreams

Not four separate jobs — one system: structure the inputs, orchestrate the week, score the workload, and explain the result.
Data Model

One JSON object, one source of truth

Structured every input — assignments, lectures, events, preferences — into a single state object per session: assignmentObjects[], lectureObjects[], eventObjects[], userPreferences{}. Every module reads from the same source instead of drifting out of sync.

Orchestration

A five-step engine, not a black box

Read deadlines → block fixed time → check capacity → distribute blocks → apply preferences. Deterministic, rule-based logic that outputs a schedule respecting real constraints — not an LLM guessing at a timetable.

Intelligence

Scoring workload before it becomes a crisis

Four heuristic scores layered on the schedule: Burnout Risk (deadline density, study gaps, workload volume → 0–100), Cognitive Load, Workload Trend, Engagement Signal — built and framed as decision-support, not a clinical tool.

AI Assistant

Claude on top, not inside, the logic

A conversational layer that reads schedule state via the Claude API and generates plain-language guidance — deliberately kept separate from the orchestration engine so the assistant explains decisions instead of making them.

04 / System

The AI decision loop

Five stages, repeated every time a student's inputs change — prediction stays separate from judgement, on purpose.
1
Data
Assignments, lectures, events, preferences ingested
2
Prediction
Burnout, cognitive load & workload scored
3
Judgement
Student reviews the read-out — not forced to accept
4
Action
Schedule adjusts once the student decides
5
Feedback
System learns the accepted preference for next cycle
Live example — burnout risk flagged high → AI analyses the week's pattern → recommends spreading study across Saturday & Sunday → projected score drops → student accepts → system learns the preference.
"AI predicts. Student decides. Always."
05 / Boundaries

My role vs. not my role

A 3-person course venture blurs job titles fast. Here's exactly where the line sat.
My role

The backend architecture and JSON data model, the orchestration engine's constraint logic, designing the four heuristic scores, integrating the Claude-powered assistant, and running the MVP validation tests.

Not my role

Pitch narrative and visual storytelling, commercial strategy and revenue modelling, and any live institutional deployment — this stays a course MVP, not a shipped university product.

Built with a 3-person team for MSIN0230 (Artificial Intelligence for Business), UCL School of Management — Aliya Abidi as Creative Lead (pitch narrative & visual design) and Shravani Naik as Commercial Co-Founder (market sizing & revenue model).
06 / Method

Before → intervention → outcome

Diagnosis → Action → ResultLMC · 2025–2026
Assignments, lectures, events & preferences lived as separate, ad hoc inputs Structured into one JSON object store — a single state per session One source of truth ▲
No consistent way to turn inputs into an actual timetable Built a 5-step rule-based orchestration engine Conflicts ▼ ~80%
Workload pressure stayed invisible until a student was already overwhelmed Layered burnout, cognitive load, workload & engagement scores on the schedule Burnout score ▼ ~53%*
Risk of an AI layer scheduling unpredictably Scoped the Claude assistant to read state and explain, not decide AI predicts, student decides

*Self-reported, 4-person pilot signal — see Validation below, not a controlled or clinical result.

07 / Product

The dashboard — everything in one place

Every component shown here is generated from the orchestration engine's output — not manually configured by the student.
Last Minute Club dashboard mockup — smart calendar, stress gauge, focus time and task load panels
CAL

Smart Calendar

Multi-week, colour-coded, conflict-free — lectures, study blocks, deadlines and events on one view.

RISK

Burnout Risk Gauge

Live 0–100 composite score across six evidence-based dimensions, always visible.

LOAD

Cognitive Load Score

Live index — deadline proximity, pending hours and active events, rolled into four states.

SIM

Trade-Off Simulator

Toggle events on or off and see the scheduling impact instantly, before committing to anything.

ASK

AI Assistant Panel

Plain-language recommendations, auto-loaded after each orchestration run.

08 / Demo

Watch it work

A walkthrough of the MVP end-to-end — inputs in, an orchestrated week and a burnout read-out out.

Last Minute Club — MVP demo.

09 / Validation

From a guess to a measured pilot

Before shipping the MVP, performance was a guess — "the orchestration engine should work." There was no structured way to check whether a generated schedule actually reduced pressure for a real student.

We ran two rounds of validation: five simulated student profiles through the orchestration engine to check generation speed and conflict reduction, and a small live pilot — three team members plus one outside student — who logged a self-reported burnout score before and after the tool planned their week.

On evidence: this is a 4-person, self-reported pilot and 5 simulated scenarios — an early signal, not a clinical or statistically powered result. Framed as such throughout.
MVP Validation LogLMC · 2025–2026
Avg. schedule generation time<25s
Scheduling conflicts reduced▼ ~80%
Aliya Abidi — burnout score88 → 42
Srishti Nisha — burnout score82 → 36
Shravani Naik — burnout score90 → 40
Fardeen Zaidi (external tester)85 → 44
10 / Evidence

Evidence & process layer

The three artefacts the MVP is actually built from.
DM

Data model & object store

assignmentObjects[] · lectureObjects[] · eventObjects[] · userPreferences{}

The single JSON state object every module reads from — designed so the calendar, the risk gauges and the assistant never fall out of sync with each other.

OE

Orchestration engine

Read deadlines → block time → check capacity → distribute → apply preferences

The deterministic, rule-based pipeline that turns raw inputs into a conflict-free multi-week schedule — no LLM in the scheduling loop itself.

AI

Assistant layer

Claude API · reads schedule state · natural-language in/out

A conversational layer bolted on top of the schedule and scores — built to explain and guide, kept out of the deterministic scheduling logic.

Product screenshots shown are from the team's own pitch deck and MVP demo — illustrative mock data, not real student records.

Field note

"Last Minute Club was a 3-person course venture, so my part sat squarely in the technical build — the data model, the orchestration engine's rules, the heuristic scoring, and wiring a Claude-powered assistant on top of it, then proving the MVP actually worked end to end."

The measurable outcomes — ~80% fewer scheduling conflicts, sub-25-second generation, and a ~53% drop in self-reported burnout scores across a 4-person pilot — came from that build. Framed honestly: early MVP signal, not institutional-scale proof.

What this shows about me

I can take an ambiguous, human problem — burnout, fragmented systems, competing deadlines — and turn it into a structured data model and a deterministic decision engine, then layer AI on top deliberately instead of by default, and validate it fast with real, if small, test data.

Where this fits

This mirrors the core of building AI products responsibly: separate what should be deterministic from what should be probabilistic, keep the human in the loop by design, and be upfront about what a small pilot can and can't prove.