Trading operations terminal

The problem

A trading operator ran dozens of bots across dozens of servers, each with bespoke parameter files and ad-hoc deployment. Strategy changes needed engineering hand-holding; on-call paging was email-based; live PnL telemetry lived in a spreadsheet that drifted from reality. Daily turnover ran into the millions but the ops team was a single person, and any production-side mistake could move money fast.

The approach

We built an internal orchestration platform with a Python strategy DSL, parameter validation at admit time, a pattern-to-strategy mapper, and live PnL telemetry over WebSocket. Redis holds the working state; Telegram delivers on-call paging with PnL context. Deploys ship to production every single day from a CI pipeline that gates on a strategy validation suite. Strategy edits ship from the operator without engineering intervention; the validation gate catches the obvious mistakes.

Stack and engineering choices

• Python strategy DSL
• Parameter validation at admit
• Pattern-to-strategy mapper
• Live PnL over WebSocket
• Redis state store
• Telegram on-call paging
• Daily production deploys

Outcome

One ops person runs the entire fleet. Strategy edits ship to production by the operator; the parameter validation catches typos and obvious mistakes before they cost money. Live PnL is the source of truth, not a spreadsheet, and on-call wakes up with PnL context already attached.