For Financial Firms

The systems small funds don't have the headcount to build.

Research, reporting, and diligence automated for independent funds and financial teams. No account manager, no six-week discovery phase. You brief the person who builds it, and something works in two weeks.

Book a call See the proof
A financial intelligence terminal showing monochrome market data with a single amber line chart

What I build

Every engagement is framed the same way: name the problem in hours and money, ship the system that removes it, measure the difference.

Deal-flow triage

Problem

Hundreds of inbound decks a quarter, read manually by the partners themselves.

Solution

An LLM pipeline that scores, tags, dedupes, and summarizes inbound against your thesis, then routes it to Notion or Affinity.

Outcome

Partners read the decks worth reading, each with a one-page memo pre-written.

Ideal: 2-5 person fund with inbox-based deal flow

Diligence research desk

Problem

Every diligence sprint rebuilds the same market map from zero.

Solution

A multi-agent research system that pulls filings, funding data, and competitor sets, and produces a structured brief.

Outcome

Multi-day market maps compress into hours.

Ideal: fund running 8+ diligence sprints a year

Portfolio & LP reporting

Problem

Quarterly LP updates are a two-week manual scramble across spreadsheets.

Solution

An automated pipeline: portfolio data in, a formatted and designed LP report out.

Outcome

Two weeks becomes an afternoon, and the report looks designed.

Ideal: fund with 10+ portfolio companies

Signals & risk monitoring

Problem

You find out about a portfolio company's problem from social media.

Solution

Continuous monitoring of filings, news, and sentiment across your portfolio and watchlist, with alerting. Risk engines (VaR, stress testing) where public exposure calls for it.

Outcome

Nothing about your positions surprises you.

Ideal: any fund holding public or late-stage exposure

Fractional AI engineering

Problem

You need engineering output, but not a hire. You have said "we should build that" at least three times this year.

Solution

Retained build capacity. You brief, I ship. Scoped monthly, cancel anytime.

Outcome

Engineering output without a salary, equity, or a hiring process.

Ideal: 2-15 person firm with a growing automation backlog

Public proof

Every claim above maps to a system you can read the source of. These are live on GitHub, not slideware.

AI Due Diligence Copilot Analyzes pitch decks, financials, and data rooms with RAG. Extracts KPIs, flags risks, drafts diligence memos. GitHub → Investment Research Assistant Automated company due diligence from live sources: markets, competitors, funding, and risk analysis. GitHub → Research Analyst Desk A four-subagent AI equity research system that collaborates to synthesize market intelligence. GitHub → Value-at-Risk Engine Historical, parametric, and Monte Carlo VaR on real market data, with stress testing. GitHub → Three-Statement Financial Model Integrated income statement, balance sheet, and cash flow modeling with multi-year forecasting. GitHub → Market Sentiment Engine FinBERT sentiment pipeline over 10-K filings, surfacing forward-looking risk signals. GitHub →

How I work

You should not have to sign a large build with someone you met online. The engagement is structured so the risk stays small until the value is proven.

Step one

Two-week paid pilot

Fixed price, fixed scope. We pick one workflow that eats your team's week and ship one working system for it. You keep everything either way.

Step two

Fixed-scope build

If the pilot earns it, we scope the full system: integrations, monitoring, documentation, and handover. Priced before work starts, not billed by the hour.

Step three

Optional retainer

Systems drift as your firm changes. A monthly retainer keeps them maintained, extended, and ahead of what you need next.

Data & confidentiality

For financial firms this is usually the first question, so here are the defaults before you have to ask.

NDA by default

Happy to sign yours before the first call. Anything you share stays inside the engagement.

Your infrastructure or mine

Systems can run entirely inside your cloud accounts. You hold the keys and the data never has to leave.

No data in third-party training

Client data is never used to train models. Where LLM APIs are involved, we use no-training endpoints and document exactly what flows where.

Deletion on request

When an engagement ends, every copy of your data I hold is deleted and confirmed in writing.

Bring me the workflow that eats your week.

Twenty minutes. You leave with a map of what's automatable and what it would reclaim, whether we work together or not.

Book a call