Substrate Build
Every team is rebuilding the same platform layer, badly.
Six to eight weeks: the multi-tenant substrate your agent systems run on — routing, memory, evaluation, and spend control — delivered to production and handed to your team.
6-8 weeks · Quoted on the call · Signed off by CTO
250-450
Concurrent agent instances delivered to production
6+
Platforms built and handed to client teams
~40%
Release-cycle overhead reduction across four teams
Who this is for
- You are shipping agents on shared infrastructure with no tenant isolation and it is starting to show.
- The platform layer is consuming the roadmap and it is not what your customers pay for.
- You need this in production in weeks, handed over, not a consulting relationship in perpetuity.
And who it is not
- You want a long-term managed platform. This is built, delivered, and handed over — the operating burden stays with your team, by design.
- You are still validating the product. This engagement is built for scale — start with the simplest version that proves the product, and bring it here when traction asks for a real platform.
What you get
- Tenant isolation and the routing layer, on AWS Bedrock, AgentCore, or Azure.
- Memory and retrieval architecture chosen against your workload, with the trade-offs written out.
- Evaluation harness and spend control built in from the start rather than retrofitted.
- Middleware and deployment templates in the languages your team actually writes.
- Handover: documentation, runbooks, and enough pairing that your team owns it before the engagement ends.
How it runs
Weeks 1-2
Architecture
Tenancy model, routing, memory, and the isolation boundary decided and written down before a line of it is built.
Weeks 3-6
Build
The substrate built against your workload, in your cloud, with your team in the pull requests rather than watching from outside.
Weeks 7-8
Production and handover
Load-tested to your concurrency target, shipped, documented, and handed over with the runbooks.
This has shipped
02
Multi-tenant agent substrate
250-450 concurrent agent instances in production. Release overhead down roughly 40%.
06
Data and ML platform standardization
One data and model lifecycle across two platforms.
08
Inference platform at scale
Cold starts 18 min to under 4. p99 from 900ms+ to 120ms. 99.9% uptime.
Read the code first
What it costs
Quoted on the call
For reference, past builds of this shape have landed in the $45-75K range over six to eight weeks. The number depends entirely on tenancy model, cloud, and how much already exists — quoting a range before seeing that would be guessing at your expense.
Every engagement can start as a $2,500 paid pilot, credited in full against the full scope if you go ahead.
Every engagement and what it costsQuestions people actually ask
Why is there no published price?
Because scope moves more here than anywhere else — a single-cloud build with an existing evaluation harness and a three-cloud build with none are not the same job. The reference range is above; the real number comes out of the scoping call.
Do we own the code?
Yes, entirely, and it is yours from the first commit. It lives in your repositories on your infrastructure throughout.
What happens after handover?
Nothing, unless you want it to. Some teams move to a fractional retainer afterwards; most do not need to, which is the point of the handover.
Fifteen minutes is enough to work out whether this is the right engagement, or whether it is one of the others, or none of them.
Book a 15-min callScoping by email works too — send what you're building and the read-back comes with a recommendation on where to start — even when the right start is smaller than you expected.