Work
The architectures we deliver today — agents, retrieval, settlement rails, market-data paths — with the measurement each one has to clear.
Each entry below is a reference architecture: the system as we build it, the evaluation it is held to, and the boundary we draw around it. They describe the work, not a client's confidential result.
The figures attached to them are illustrative until a client approves the real ones, and they are marked as such. We would rather show you an honest architecture than a fabricated number.
Dealt in order. Open any card for the architecture, the evaluation and the boundary.
A permissioned agent reads intake documents, drafts adjudication, and escalates anything below a confidence floor. Every action is traced and replayable.
Normalised multi-venue ingestion with deterministic replay, so teams can reproduce any minute of the day. Infrastructure only — no strategy, no advice, no return claims.
OCR repair, entity resolution and a domain fine-tune turned an unsearchable corpus into a retrieval layer with a defensible accuracy benchmark.
Contracts and an indexer with a full replay suite, so every state transition is reproducible before it ever reaches mainnet.
A permissioned agent reads intake documents, drafts adjudication, and escalates anything below a confidence floor. Every action is traced and replayable.
Architecture: Graph + sandbox · P95 latency: 1.4s / step · Eval coverage: 312 cases
Normalised multi-venue ingestion with deterministic replay, so teams can reproduce any minute of the day. Infrastructure only — no strategy, no advice, no return claims.
Throughput: 1.2M msg/s · P99 ingest: 840µs · Replay: Bit-exact 90d
OCR repair, entity resolution and a domain fine-tune turned an unsearchable corpus into a retrieval layer with a defensible accuracy benchmark.
Pipeline: 7 stages · Retrieval F1: 0.91 / 0.63 · Corpus: 2.1M docs
Contracts and an indexer with a full replay suite, so every state transition is reproducible before it ever reaches mainnet.
Coverage: 100% branch · Audit findings: 0 critical · Indexer lag: < 1 block
Tell us the system you can't get built. We come back with a short, paid discovery — a clear plan and a fixed first milestone — usually within two working days.
A working session to map the problem and define what "good" is measured against.
Architecture, milestones and a fixed first deliverable — yours to keep, either way.
Embedded with your team or as a dedicated pod, shipping with traces, evals and docs.