Skip to content
Services

We are hired for the difficult half of the programme.

Advice you cannot implement is decoration. Four things, each of which we have built and can show you, and nothing we would have to learn on your budget.

Engagement shapes
Architecture review: 2–4 weeks, fixed priceThin slice to production: 6–12 weeksEmbedded delivery: Quarterly, cappedFractional architecture: Retained, a few days a month
Commercials
Fixed price for the review. Everything after that is capped time-and-materials with a named team and a monthly exit point.

AI & Agent Engineering

The layer that makes an agent safe to leave running: an explicit orchestration graph, evaluation that gates deploys, capability scopes enforced at call time, and cost you can attribute per task.

  • Orchestration design

    Task decomposition, tool contracts, retrieval and memory, and a graph that a long-running agent can be resumed from rather than restarted.

  • Evaluation harnesses

    Golden sets, scoring rubrics and regression gates in CI, so a prompt change cannot quietly degrade a path nobody read that week.

  • Capability scoping

    Tools behind grants checked at call time instead of asked for in a prompt, with human confirmation on anything irreversible.

  • Model routing & cost

    The cheapest adequate model per hop, hard budget ceilings, and cost attributed per task rather than per invoice.

Read the practice detail

Platform & Systems Engineering

Rust and WASM systems work, plugin boundaries, distributed job scheduling and the deployment shells around them. These are the parts that decide whether any of it survives contact with production.

  • Plugin architecture

    WASM component boundaries, host-call surfaces and versioned interfaces, so extensions cannot reach past the one door you gave them.

  • Ports and adapters

    Domain logic that depends on traits, with each environment supplying its own adapter, so a new backend becomes a new adapter rather than a rewrite.

  • Distributed execution

    Peer discovery, leases, fairness and warm pools for scheduling work across machines, including intermittently reachable ones.

  • Deployment shells

    The same core shipped in-process, as a daemon, or compiled to wasm32, chosen per target rather than per rewrite.

Read the practice detail

Product Engineering

Design and full-stack delivery in one small squad, shipping to real users in increments, with your engineers in the pairing rotation from the start.

  • Discovery & framing

    Time with the people doing the work, turned into a problem statement someone can actually fund.

  • Interface & design system

    Accessible components documented as living code rather than as a slide library.

  • Full-stack delivery

    React, Rust, Tauri, Node and the edge, shipped to production early and often rather than demonstrated at the end.

  • Hand-over

    Documentation written, tests green, deployment path owned by your team before we leave.

Read the practice detail

Architecture Review

A short, fixed-price read on a system or an in-flight programme: what will break, what it will cost, and what to do about it in what order.

  • Assessment

    Reading the code, the data flows and the deployment path, and talking to the people who carry the pager.

  • Threat modelling

    Trust boundaries and agent capability scopes analysed for the ways they actually get abused.

  • Cost modelling

    Unit cost per transaction at today's volume and at ten times it, with egress and idle capacity made visible.

  • Sequenced options

    A ranked plan with owners, effort and the conditions under which we would reverse each decision.

Read the practice detail
Not sure which one

Most problems arrive mislabelled.

Describe the symptom and we will tell you which practice owns it, or that you do not need a consultancy for this at all.