Enterprise engineering
Every engagement ends in something operable: a costed sequence, a monitored system, an evaluation harness you keep. No capability decks.
Six engagements
S-01
A four-week diagnostic: where your data actually lives, which use cases clear a payback threshold, what has to be fixed before a model touches anything. Ends with a costed sequence, not a slide deck.
Deliverables
S-02
Retrieval that survives real corpora: hybrid search, structure-aware chunking, permission filtering, freshness guarantees and citation enforcement. The unglamorous work that separates a demo from a system.
Deliverables
S-03
End-to-end automation of document-heavy processes — intake, extraction, validation, exception routing, system-of-record write-back — with exception rates you monitor rather than discover.
Deliverables
S-04
Before go-live: golden datasets, adversarial suites, PII handling, refusal behaviour, rollback plan. After go-live: drift alerts, cost telemetry and an evidence trail.
Deliverables
S-05
When data cannot leave the site: model distillation, quantisation, board selection or custom Xelec design, thermal and power validation, and fleet management for devices in the field.
Deliverables
S-06
Risk classification, technical documentation, human-oversight design, logging and transparency obligations — mapped to the systems you are actually running, with your legal team in the room.
Deliverables
Also available
Hardware engineering from Xelec.
IoT systems, embedded prototyping, firmware development and full electronics product development — run with or without an AI component.
Mapping
| Symptom | What we do about it |
|---|---|
P-01 The pilot graveyard | Paid pilots with a written production gate — pass and we ship it, fail and you keep the findings. |
P-02 The data is trapped in documents | Document intelligence and a data foundation layer built before any model work starts. |
P-03 Compliance is the real blocker | EU and US residency options, model-choice control, DPIA-ready documentation and full audit trails from day one. |
P-04 Nobody can prove it is right | Evaluation sets, citation-grounded answers, guardrails and drift monitoring shipped with the system. |
P-05 Vendor sprawl and unpredictable spend | Model-agnostic architecture, cost telemetry per workflow, and the option to move to smaller in-house models. |
P-06 Cloud-only AI cannot reach the floor | On-device inference on Xelec hardware — the capability most AI vendors simply do not have. |
How we contract
Discovery credits against a pilot. Pilots are fixed fee. Nothing open-ended until you have seen us work.
Length: 2 weeks
Commercials: Fixed fee
Use-case scoring, data reality check, costed sequence. Credited against a pilot if you proceed.
You have candidate use cases and no agreed sequence.
Length: 4–8 weeks
Commercials: Fixed scope, fixed fee
One use case taken to a written production gate, with the evaluation harness you keep either way.
You need proof that survives an operations review.
Length: Quarterly
Commercials: Embedded team
A senior team embedded against a roadmap: production systems, hand-off runbooks, internal capability transfer.
You are building an AI portfolio, not one feature.
Length: Annual
Commercials: SLA-based
We run and monitor what we built — drift, cost, guardrails, upgrades — against an agreed SLA.
You want the system owned, not just delivered.
Next step
We will tell you which engagement fits, or that we are the wrong partner. Both answers are useful.