Deterministic where it matters
The model reads messy input. It does not get to be the judge. Scoring runs on a rule pack you can replay and retune.
ColdDeck scores prospects with a rule pack, explicitly not a model.Production AI Engineering
production.
Every output traces to the evidence that produced it. A human approves anything irreversible. The pipeline heals itself when it breaks. Because “the model said so” is not an answer your customers will accept.
These four things show up in every system we have shipped, across two clients and three products. They were never planned as a brand. They are just how the work got done — which is what makes them worth reading.
The model reads messy input. It does not get to be the judge. Scoring runs on a rule pack you can replay and retune.
ColdDeck scores prospects with a rule pack, explicitly not a model.A number you cannot point at is not an answer. Every record carries the page, the timestamp, or the frame it came from.
Every score point traces back to the signal that produced it.Nothing that cannot be undone happens unsupervised. The system drafts; a person approves.
ColdDeck drafts every email and waits for approval before it sends.Queues stall, machines sleep, APIs rate-limit. The pipeline is expected to run unattended for months and notice its own gaps.
DripDeck reassigns missed slots so the queue never silently stalls.cold email that reads like you actually looked.
Finds businesses in a niche, crawls each site for concrete defects, and scores them on a deterministic rule pack — explicitly not a model. Every point traces to the page that produced it. A human approves every email before it sends.
Measured
two clocks, and neither one needs you awake.
Uploads early and private with a publishAt timestamp, so the platform itself flips the video public at the right moment. Days the machine was off self-heal — the next run notices the missed slot and reassigns it.
Measured
Two further systems — multimodal extraction with end-to-end provenance, and document extraction with auditable confidence scoring — are client work and are not published.