I build data and AI products for private capital, end to end.
Private capital is money invested in companies and loans that don't trade on a stock exchange. These investors decide with little public data. I turn their messy documents into data they can trust, then into tools that speed up their decisions.
| Product | Example Persona | Example Private capital use case | Job to be done | Result |
|---|---|---|---|---|
| Benchmarking product (BlackRock and Preqin), first client-facing end-to-end version | Deal team at a private equity firm | Comparing a company they want to buy with similar past deals | When I price an acquisition, I want to see what similar companies sold for and how those deals performed, so I can set my bid and test my return assumptions | Better deal pricing, faster. More deals, better returns |
| Loan data from 6,000+ unstructured BDC filings | Deal team at private credit firm | Portfolio Monitoring, Portfolio Protection | Identifying segment of markers showing distress, and whether the portfolio is exposed | Increased and timelier visibility on Private Credit Distress, reducing likelihood of loans defaulting |
- Use AI where it beats fixed rules, not by default.
- Define what "correct" means before measuring quality.
- Make every output traceable to its source. If users can't check it, they won't trust it.
- Treat monitoring and governance as part of the product.
Prototypes and proofs of concept, built outside work on public or made-up data so anyone can try them. My main work is at BlackRock, with real data for real clients, and isn't shareable.
See all six with case studies and demos →
| Project | Who it's for | What it does | |
|---|---|---|---|
| BDC Footnote Tape | Analyst at a private lender | Turns any lending fund's SEC report into one row per loan, with its footnotes explained | Case study · Demo |
| BDC Mark Tape | Lender pricing or reviewing loans | Shows what other lenders charge and how they value similar loans, with sources | Case study · Demo |
| Deck Rank | Junior investor at a startup fund | Ranks incoming pitch decks so the best is read first; every score quotes its slide | Case study · Demo |
| Alpha Scout | Junior investor at a startup fund | Finds startups before they ask for money; every score has a quote and a link | Case study · Demo |
| Night’s Watch | Credit team at a private lender | Warns when a borrower may not have the usable cash for its next interest payment | Case study · Prototype · Deck |
| Shield | Fund that owns part of a construction project | Finds contract deadlines in project documents before the right to claim money expires | Case study · Demo |
I run product work through PM OS. It started from Aakash Gupta's PM OS, which I customised into my own set of 48 product skills in Claude Code and Devin. It takes an idea from opportunity check to PRD, review by 7 AI reviewers, build and results. Each correction I make becomes a rule. AI does the repetitive work. I decide what matters and what ships.
Tools: Claude Code, Devin, Python, Streamlit, LangFuse (AI monitoring), DoWhy (causal inference), dbt, Snowflake.
When I'm not building, I'm working on my backhand, in the gym, or out on a walk.