Data that survives a bad night
Normalised feeds, gap detection and replayable archives. If a venue drops a session, you should know before your model does.
Independent engineering practice
Infrastructure for teams that trade their own models.
We write the unglamorous parts: tick collectors that do not lose data, backtests that match production, and order routers that fail loudly instead of quietly.
What we usually get called for
Backtests you can argue with
Event-driven engines with explicit fees, latency and partial fills, so a result can be defended rather than merely presented.
Execution with brakes
Position limits, kill switches and reconciliation against the venue, wired in from the first commit instead of after the first incident.
How the work runs
- Week one
- We read the existing system and write down what actually happens, including the parts nobody documented.
- Weeks two to six
- Short iterations against a staging venue or a simulator, with a written note at the end of each one.
- Handover
- Runbooks, dashboards and a deployment your own engineers can operate without us.
Availability
We take a small number of engagements at a time and are usually booked several weeks out. Enquiries go to hello at tradecode dot tech.