Audit work comes in two shapes. Either you have a system you rely on and want an outside engineer to look for what you have stopped seeing, or someone is selling you an algorithm and you would like it examined before you commit capital.
What we examine
- Correctness: does the code do what the specification says, in every branch
- Look-ahead and data leakage in the research pipeline
- Cost and fill assumptions versus reality
- Risk controls: caps, stops, kill switch, restart behaviour
- Credential and key handling, logging hygiene, dependency risk
- Reproducibility: can the reported numbers be regenerated from the repository
On vendor claims
If a vendor cannot supply a trade-by-trade log, the cost assumptions and an out-of-sample window, there is nothing to verify, and we will say exactly that. We assess evidence and engineering quality. We do not endorse products, and we never conclude that any algorithm will be profitable.