I build quantitative infrastructure at fixed price and fixed date, and every build ships with a validation report that says where it is wrong. If it is late, or the report is missing, you do not pay.

Two things I shipped last week, alone, in two days each — both public, both still running:

  • tirramind-universe — every US delisting since 2017 reconciled from SEC filings with the cause of each, plus 37,651 insider sale notices linked to the Form 4 sales that executed them. Hand-checked 29/30 against EDGAR. Refreshes itself daily.
  • voltorch — an arbitrage-free volatility surface for Deribit BTC/ETH, refit every 30 minutes, with its fit error inside the spread published per expiry. Three numerical bugs found in the process, each pinned by a test. Live page.

Offers

1. Vol surface & greeks engine for your desk

An arbitrage-free implied-volatility surface on your venue and your data — eSSVI backbone that cannot contain butterfly or calendar arbitrage, per-expiry refinement checked on a dense grid, greeks with error bounds against closed form, and an executable-arbitrage monitor on the live book.

In scope: one venue (Deribit, Bybit, OKX, CME, NSE), one asset class, bid/ask implied-vol extraction with a solver that cannot stall, fit report per expiry, Python package + CLI + optional REST endpoint, deployment on your box or mine, 30 days of fixes. Out of scope: execution, position management, market data licences, American exercise (available as an add-on).

$8,000–15,000 · 2 weeks. India: ₹6–12 lakh.

2. Model validation certificate

You have strategies — yours, or the pile an AI wrote — and someone wants a number they can trust. I run walk-forward validation with future-blind windows, interval coverage tests, calibration slope, deflated Sharpe, multiple-comparisons control, and effective sample size with cluster bootstrap. You get a report with every number carrying its uncertainty and a one-page certificate you can put in front of an allocator or a regulator.

In scope: up to 50 strategies or one model, your data, a reproducible notebook, a written report, a call to walk it through. Out of scope: fixing the strategies; telling you what you want to hear. Roughly a third of what I have measured in my own work came out negative and was published that way.

$3,000–8,000 · 1 week. India: ₹2.5–6 lakh.

3. Regulatory-filing ledger

Any regulator’s filings — SEC, SEBI, FCA, ESMA, an exchange’s circulars — reconstructed into a queryable, evidence-linked dataset that refreshes itself: every row links to the filing that proves it, every unknown is reported as unknown, and a hand-checked sample ships with it.

In scope: one regulator or venue, up to ten form types, daily refresh job, parquet + API, a published disagreement ledger against whatever list you use today. Out of scope: paid data feeds; anything the source’s terms forbid.

$5,000–10,000 · 2 weeks. India: ₹4–8 lakh.

After the build

Daily jobs, monitoring, changes and questions: $2,000–5,000 / month, cancel any time.

How it works

  1. Send a paragraph about what you need — email or X. I reply within a day with a fixed price, a fixed date, and what “done” means.
  2. Half on start, half on delivery. Invoice or card.
  3. You get the code, the report, and a repository you own.

I work alone, with heavy use of AI tooling, which is why it is fast. The validation report is why it is safe.