How I work

Three phases

Each phase ends with something running. If a phase does not produce a working piece, it is not finished.

• 01 · Research

  Data ingestion and storage, then a backtesting environment. You can test a strategy yourself, on your own data, before any money moves. This is the phase that proves fit. Available on its own, including an audit and rebuild for teams that already trade.

• 02 · Execution

  Order management and broker connectivity over the broker's API or FIX. An order reaches the market, comes back filled, and the position is where it should be.

• 03 · Risk and reporting

  Position, P&L, limits and a daily report. The desk trades inside its limits and leaves a record of what it did.

How the work runs

• Cadence

  A working demo every two weeks. You see the desk, not a summary of it.

• Timeline

  A Core Desk takes up to three months from agreement to first live trade. Larger desks are quoted with their own timeline.

• Your server

  A shared machine on your cloud or hardware, reached over VPN. The team works in JupyterLab on the same data, runs backtests there, and scheduled jobs run on it. Documented and reproducible from a clean machine, like the rest of the desk.

• Billing

  Fixed fee per scope, billed against milestones. No hourly rate, no timesheets.

• Decisions

  One decision-maker on your side. Ambiguity about who approves is the most common reason a build slows down.

• Code I bring

  Components I reuse are licensed to you rather than sold. The terms are agreed in writing before work starts.

What happens if I disappear

You are hiring one person, so ask this first. It deserves an answer you can check rather than reassurance.

The short version: the desk keeps trading without me. New work stops until someone else picks it up, and everything they need is already yours.

• Source

  Source for each milestone is in your repository once that milestone is paid. Nothing is held back until handover.

• Access

  Servers, cloud accounts, broker keys and data subscriptions are in your name. I work with access you grant and can revoke.

• Tests

  Behaviour is pinned down by tests, so someone else can change it safely.

• Runbooks

  Every operational task written down: deploy, restart, reconcile, recover. Your team runs them without me.

• Standard parts

  FIX, Python, SQL, Linux and ordinary databases. Any competent engineer can pick it up; there is no proprietary runtime to depend on.

• Reused code

  Components I bring are licensed to you perpetually. The licence does not end if I do.

• Planned exit

  A walkthrough with whoever inherits the desk. Terms are set in the contract.

There is no bench behind me. If you need a team with an on-call rota, hire a team. What you get from me is one person who has built the whole pipeline several times, and a desk that does not need me to keep running.

The desk I run

I run my own book on the same components I build client desks from: data, research, signal generation, execution and Monte-Carlo risk simulation, developed since 2017. It is my own capital and my own infrastructure, so it is evidence of the method rather than of client delivery.

I don’t publish returns, track records or performance claims. The desk is here to show how the machine is assembled and that it survives contact with a live market, not to sell a strategy.

See the projects →

← Back to home