I build the whole function: financial planning and the operating model, fundraising and investor relations, capital structure, board reporting, and the unit economics the business runs on. Plus the AI and automation layer that lets a small team operate like a much larger one.
That is the job I do. The proof is a deterministic finance system I designed and operate, running a paying startup's books in daily production since early 2026. Every number computed in code; a human owns the review.
Every financial transaction assembles into one structured Register. From there, role-gated views render automatically for the people who need them: the operator sees one view, the board sees another, all from the same source. This is the control surface of a system running in production today. The data below is illustrative.
| Date | Customer | Project | RevType | Amount | Status |
|---|---|---|---|---|---|
| 2026-01 | Acme Industries | Q1 Implementation | Subs | 12,500 | Active |
| 2026-01 | Beta Corp | Onboarding | Advisory | 8,250 | Active |
| 2026-02 | Cipher Co | Phase 2 | Subs | 15,000 | Active |
| 2026-02 | Delta Holdings | Custom Build | Implementation | 22,500 | Pipeline |
| 2026-03 | Echo Systems | Q1 Support | Subs | 9,800 | Pipeline |
| 2026-04 | Foxtrot Labs | Phase 3 | Subs | 18,000 | Provisional |
| 2026-04 | Gamma Holdings | Custom | Advisory | 6,750 | Pipeline |
One structured event log of every financial transaction. Each row carries date, customer, project, type, amount, and status. Historical, forecasted, and in-pipeline rows live together, kept current in code with a human owning the review.
| Mar | Apr | May | Jun | |
|---|---|---|---|---|
| Revenue | 24K | 28K | 31K | 35K |
| COGS | (8K) | (9K) | (10K) | (11K) |
| Gross Profit | 16K | 19K | 21K | 24K |
| OpEx | (12K) | (14K) | (15K) | (16K) |
| EBITDA | 4K | 5K | 6K | 8K |
Cash-basis P&L that reshapes per scenario in one click.
| Customer | FY26 |
|---|---|
| Acme Industries | 156,000 |
| Beta Corp | 98,500 |
| Cipher Co | 124,000 |
| Delta Holdings | 89,300 |
| Echo Systems | 67,800 |
Customer rollup with FY totals and recognition status.
| Vendor | FY26 |
|---|---|
| Cloud Services | (24,000) |
| Marketing Co | (45,000) |
| Office Supplies | (8,400) |
| Payroll | (380,000) |
| Software | (18,500) |
Vendor rollup by category. Payroll separate. Accruals reconciled.
| Invoice | Term | Recognized |
|---|---|---|
| INV-1024 | 12 mo | 3,125 |
| INV-1025 | 12 mo | 1,375 |
| INV-1026 | 6 mo | 2,500 |
| INV-1027 | 12 mo | 7,500 |
| INV-1028 | 6 mo | 4,000 |
Deferred-to-recognized waterfall, computed from each invoice's term and start date.
Most early-stage founders are the human API between their books, their billing, their bank, and their own head. Pull from one, paste into another, translate for the team, update the forecast. Every handoff loses context. Every spreadsheet is a single point of failure.
This is the function I build, and the system that runs it.
It connects to the tools you already use - general ledger, payroll, banking, CRM. No rip-and-replace.
Last month, on a paying startup's live model, I worked ten change requests through the system end-to-end: replies drafted, figures updated, the model refreshed, every number computed in code with a human owning the review.
Every founder has tried ChatGPT on their numbers. Here's why it didn't stick.
Every calculation runs deterministically in code against the company's general ledger. No guessed numbers.
The model keeps its context across the engagement, so the work builds on what came before instead of starting over.
A change in the pipeline flows through to cash, runway, and hiring, because it all lives in one connected model.
Cai stays in context across the artifacts your team already uses - the spreadsheet, the deck, the doc, the chat channel. One conversation, one memory, one source of truth.
One system, built around how the business actually runs. Here is what standing it up looks like.
General ledger, payroll, banking, CRM. No rip-and-replace.
A single Register of every transaction, rendered into role-gated views built around how the company works.
The model stays current in the spreadsheet, deck, doc, and chat your team already uses.
I operate this system for one founding client today, and I am looking for the right next team to build it into - as an operator, or alongside one. Talk to Cai or email Ty.
AI runs on your own subscription and API keys with zero markup on your compute.
Your context, your data, your outputs - portable and yours to keep. The methodology is licensed; the work product is owned.
Every number is computed in code and traces back to source, reviewed by a human before it reaches the board.
A decade in institutional finance: investment banking at Houlihan Lokey and William Blair, a relative-value hedge fund, two startups co-founded, and years as a startup finance operator. Builds the whole strategic finance function from the ground up, and architects the AI system that runs it.
The system Ty designed and operates. It assembles the finance model, renders the role-gated views, and keeps them current. Ty directs the work and a human owns the review; the math runs in code.
Ty built the methodology and operates the system. The track record is the spine; the system is the multiplier.