Kumu / Corporate Finance

International Corporate Finance

The Performance Ratios project — AI + GitHub throughout: build a public portfolio repo, model a real company's financials, specify an analysis precisely enough for an AI to execute it — then judge whether the AI got it right. Then value the company: adapt a professional DCF skill, run it, audit it, and take a position against the market price.

Graduate · 9 stages Deliverable-only — no presentations

Stage 0 — get GitHub working, before the project

Every stage of this project is handed in by committing it to a public repository of your own, so the repository comes first — before the template, before the company memo, before any analysis. One sitting, once, and the rest of the term is about finance instead of tooling.

It comes in two parts: the account, then the workspace.

1

Part 1 — Get a GitHub account never used GitHub? start here

Create the account and learn the three verbs: add → commit → push. If you already have a GitHub account, use it and go straight to part 2. Any email works for your GitHub account; for the free GitHub Education benefits, add your school address under Settings › Emails and apply, which is optional and never graded.

account + mechanics
2

Part 2 — Set up your workspace

Directory skeleton, plus a bio and resume you write yourself. AI may review them, never draft them.

Repo URL

Part 1 in full: the Git mechanics walkthrough

Start Here → Git mechanics walks the whole workflow once: local versus remote, add → commit → push in the desktop app and on the command line, editing files straight on github.com, and fixes for the mistakes everyone makes the first time. Read it before part 2 — the stage page assumes you can already save work to a repo, and does not re-teach the mechanics.

0 of 9 stages complete

The learning path — stages 1 to 8

Each stage produces a named artifact the next stage consumes. Work them in order — the chain is the point.

1

Provided Ratios Template

Understand the provided Excel architecture: tabs, color coding, and the named-range system everything else depends on. Install the performance-ratios skill.

.xlsx + skill
2

Company Selection Memo

Pick a listed company with a share price you can cite, defend the choice like a senior analyst, and set falsifiable hypotheses. Grant instructor access.

.md memo
3

Populated Financials

Enter two years of real statements, tie every number to the source filing, run the skill's identity checks, and watch the Ratios tab come alive.

.xlsx build
4

Technical Specification

The central artifact: draft a spec yourself, then review and refine it with an LLM, until it is so precise that another AI — with zero context — can execute your analysis.

.md spec
5

LLM Ratio Analysis & Executive Evaluation

Run your spec through an LLM, verify its numbers by hand, write the executive judgment only you can provide.

.md analysis
6

Modify the DCF Skill

Copy a professional DCF skill in unmodified, then edit it until it fits your company's currency, workbook and capital structure. The annotated diff is the spec.

skill + diff
7

Run & Audit the DCF

A fresh agent session builds the model from your skill. Audit it twice: with a second professional skill, and by hand.

.xlsx + audit
8

Valuation Memo & Repo Polish

The capstone: a board memo setting your value per share against the market price, then a repo that stands as a portfolio piece.

Repo URL

Practice first

Interactive lab

Ratio Lab

Two years of statements in, all six ratio categories out — with the balance-sheet check, startYear_* averaging, and a Du Pont decomposition validated against direct ROE. Mirrors performance-ratios-template.xlsx.

Field guide

AI Tools Lab

The field guide to coding LLMs: chat vs "code" tools, web vs desktop vs CLI, and why giving the AI your repo is the single biggest upgrade. Read it before your first AI build session.

The thread that ties it together

Stage 0 ── set up your public portfolio repo
   ↓
Stage 1 ── upload the ratios Excel template          models/templates/
   ↓
Stage 2 ── memo: choose & justify your company       docs/decisions/
   ↓            (instructor reviews via pull request)
Stage 3 ── populate template with real financials    models/builds/
   ↓
Stage 4 ── you draft the spec; an LLM reviews it     capabilities/performance-ratios/
   ↓            (one human-in-the-loop iteration, documented)
Stage 5 ── feed ONLY the spec to an LLM → verify,
   ↓       evaluate, correct → final analysis        analysis/
Stage 6 ── copy a DCF skill in, then adapt it;
   ↓       annotate every edit                       .claude/skills/, capabilities/valuation/
Stage 7 ── a fresh session runs the skill;
   ↓       audit by tool and by hand                 capabilities/valuation/model.xlsx
Stage 8 ── valuation memo against the market price   docs/decisions/
           polish the repo → submit the URL

The finished repo holds the whole chain: the template and its skill, the memo, the populated model, the spec, the analysis, the modified DCF skill, the valuation model, and the prompt log.

Skills the project practices

SkillWhere it shows up
AI literacy & prompt engineeringYou draft, an LLM reviews, then you iterate together at Stages 2 and 4; your Stage 0 bio is your own draft, which AI may review but never write; evaluating LLM output at Stage 5; adapting and auditing an agent skill at Stages 6 and 7; the prompt log throughout
Spec-driven designStage 4's core lesson: specifying work precisely is more valuable than executing it
GitHub & version controlEvery stage commits to your repo; PR-based feedback from Stage 2; skills versioned as a baseline plus edits at Stage 6
Financial modelingNamed-range architecture (Stage 1), real-data population, tie-out and identity checks (Stage 3)
ValuationStages 6 to 8: free cash flow, discount rate and terminal value in one currency, the bridge from enterprise value to a value per share, and a beta × terminal growth sensitivity table
Executive judgmentStage 5's verification table, LLM critique, and the investment thesis in your own voice; Stage 8's verdict against the market price

House rules

Weights and due dates are in your course schedule.