Selected work
01 · Public-source financial planning
Strategic Finance & Corporate FP&A Model
A source-grounded scenario workflow that connects AI-cloud demand, capacity expansion, capital needs, and liquidity.
SEC EDGAR→XBRL→Source ledger→Scenario layer
Executive summary
A portfolio finance model organized around the management question: can contracted AI-cloud demand convert into revenue while funding infrastructure expansion without unacceptable liquidity or leverage risk?
Problem
Public disclosures can be fragmented across filings and investor materials. The project creates a traceable path from reported information to a structured operating and financing view.
Approach
- 01Discover and cache filings, extract XBRL facts, and record source metadata in a source ledger.
- 02Roll committed demand through an RPO bridge, capacity proxy, revenue, capex, debt, interest, and cash flow layers.
- 03Run base, upside, and downside cases with assumptions explicitly labeled rather than presented as reported actuals.
- 04Export a formula-based workbook, processed datasets, validation checks, and a board-style presentation package.
02
Data & methodology
Data sources
- SEC EDGAR filings and XBRL company facts
- Official investor-relations releases and presentations when available
- Clearly labeled analytical assumptions and proxies
Methods
- RPO roll-forward and scenario-specific conversion curves
- Capacity Deployment Index and Revenue Yield Proxy when public operating data is incomplete
- PPE, capex, debt, interest, working-capital, and liquidity schedules
- Forecast-versus-actual variance workflow with inference labeled as inference
Workflow
Input→Collection→Analysis→Output
Public filings and official investor materials; every non-reported driver is labeled in the model.
Outputs & findings
Key outputs
- Formula-based strategic finance workbook
- Board-style PowerPoint pack
- SQLite data store, processed tables, and validation checks
- Streamlit interface for scenario review
What the work demonstrates
- Created a repeatable, source-led planning workflow rather than a point estimate.
- Maintained an explicit distinction between reported actuals, derived figures, assumptions, and analytical proxies.
Limitations
- No confidential contracts, exact GPU utilization, site-level schedules, or per-GPU pricing.
- Capacity and revenue-yield measures can be analytical proxies when direct public metrics are unavailable.
- Educational portfolio work only; not investment advice or CoreWeave internal planning.