Crypto trading bot development and the machinery underneath a trading operation: the execution path, the simulator it gets tested in, the limits that stop it, and the settlement and reconciliation that establish what actually happened.
Our job is making sure the system underneath does what you believe it does.
The failures that cost money are quiet. A fill the backtest assumed and the venue never gave.
A socket that reconnected cleanly and resumed from a stale sequence number, so the book was wrong for a while and nothing logged it.
An exposure limit that lives in a spreadsheet somebody checks after the close. A ledger that agrees with itself, disagrees with the exchange by a handful of trades, and is discovered at month end.
Most of that is preventable by refusing to let the test environment be a different program.
The backtester, the simulator and the live path run the same execution code with a different data source and a different clock. That is slower to build.
It is also the only arrangement where a passing test tells you anything, because a venue adapter that exists only in the live path is a component nobody has ever tested.
Anupam Shah ran payment infrastructure at $60M+ volume in MAS-regulated, SOC 2 / ISO 27001 environments.
That is where the settlement, reconciliation and audit-trail half of this comes from.
Rust where the correctness or latency argument survives being questioned, and something duller where it does not.
Most of a trading stack is not the hot path, and rewriting all of it in Rust is a way to spend a budget rather than a way to be right.
You own the edge and the outcome. We design no strategies, take no discretion over your capital, and publish no performance figures.
Scope
Assumed fills, state that went stale during a reconnect, limits enforced outside the code path, breaks that only appear at reconciliation. We walk the system you already have and write the list down.
The list is the scope, and some of it is cheaper to fix than to rebuild.
Venue and protocol adapters written against observed behaviour rather than against the documentation, and simulation against forked state so a change can be argued about before it touches capital. The simulator runs the production code, not a friendlier copy of it.
Position, exposure and drawdown limits enforced by the code path instead of by discipline. A breach halts rather than warns. The kill switch gets tested by pulling it in the simulator, not by being written down.
Reconciliation across your own ledger, the venue and, where it applies, the chain, so a break surfaces on the day it happens rather than at month end. An audit trail that still reads clearly when the person reading it is not you.
Deterministic backtests, pinned data, a test suite your own engineers can run, and a written runbook with a rollback path. A result you cannot reproduce is not a result, and a system only we can operate is a dependency we would not accept either.
A fixed fee against written scope, agreed before work starts and billed at milestones as the work lands.
You do, outright. IP assignment is signed before repository access, by everyone who touches the code, and it covers AI-assisted output. Handover includes the runbook and the rollback path. Nothing depends on a licence we control.
Nobody before an NDA. Access is read-only until scope is signed, and you can revoke it at any time. The AI tools used on your code, and what each one retains, are listed on the governance page.
Founder, Provingmark
Thirty minutes, no charge.
Related: DeFi protocol engineering · AI agents for on-chain systems · About Anupam Shah · How we work