OnPoint Technologies, LLC We build with founders

We build with founders. Not for clients.

Most startups hit a wall because the foundation they shipped at speed can't carry what comes next. We take the CTO seat at early companies, build the layer underneath, and take a stake in what gets built. That foundation now includes AI. Build without it and you ship something already dated.

Platform
.NET 10 · ASP.NET Core
Data
SQL Server 2025
Cloud
Microsoft Azure
Role
Fractional CTO — in the seat, in the repo
AI
Designed in, not bolted on
Position
Equity, not invoices
months years decades bedrock INTERFACE screens · reports · API surface SERVICES workflow · rules · integration DOMAIN MODEL entities · invariants · language DATA FOUNDATION schema · keys · constraints AI
Fig. 1 — A system in section. The top layer is rewritten every
few years; the bottom is what companies rewrite at the worst
possible moment, or never have to. AI cuts through all four —
it is not a layer you add on top.
§ 01

We're in it, or we're not in it.

We don't take briefs and hand back a codebase. We join a small number of companies a year — in the CTO seat, in the repository, and on the cap table.

Leadership

We take the CTO seat

Fractional CTO for companies that need senior technical judgment now and can't yet hire it full time. You get the seat filled, not a report about filling it.

  • AThe architecture calls that are expensive to reverse
  • BHiring engineers, and knowing who is actually good
  • CThe technical answer in the room when investors ask
Engineering

And we build it too

Leadership that never touches the code goes stale within a quarter. We stay in the repository, so the plan and the thing being built remain the same thing.

  • AWe join early, while foundations are still cheap to get right
  • BWe sit in the product decisions, not just the ticket queue
  • CWe hold a stake, so we're paid when the company works
Our own

Sometimes the founder is us

When we see a problem worth solving and nobody is solving it well, we start the company ourselves and run it as a product.

  • ACommercial software in production with paying customers
  • BCompliance and document-heavy markets others avoid
  • CWhere the standards we bring to partners came from
§ 02

AI isn't a feature. It's foundation.

A product designed without AI at its core is already behind on the day it launches. We build it into the model from the start — and we use AI to do the building, which is how a small team delivers like a larger one.

In the product

Designed in from the schema up

AI capability belongs in the data model, not appended to a finished product. What it reads, what it decides, what it must never guess — and which model does each job — are foundation decisions, made before the first screen exists.

  • ADocument understanding and structured extraction
  • BClassification and matching against reference data
  • CModels picked per task, not one vendor stretched across everything
  • DEvery result sourced, checked, and correctable by a human
In how we build

A small team that delivers like a larger one

AI carries the mechanical work — scaffolding, schema, tests, first-pass review — so senior judgment goes where it compounds. For a company we hold a stake in, that difference is runway.

  • AAnthropic’s Claude as our development AI, reviewed line by line
  • BTests written alongside the code rather than promised for later
  • CHuman judgment on the parts AI gets confidently wrong
§ 03

What we have a stake in.

Companies we co-founded, companies we build with, and products we run ourselves.

Co-founded OnPoint Technologies Financial services. Buy-side company analytics. Acquired · S&P
Co-founded Prima Capital Financial services. Investment research, analytics, and portfolio modeling. Acquired · Matrix
PE turnaround Corptax Hired as CTO for this Mason Slaine/Warburg Pincus turnaround. Completed
Building with Accountable Africa AI An early company where we hold the engineering side and a stake. Early stage
§ 04

Narrow on purpose.

One ecosystem, known end to end. A company we back doesn't get whatever is fashionable — it gets a stack we can still operate at scale in five years.

LayerWhat we use
Runtime.NET 10 — ASP.NET Core, MVC, minimal APIs
DataSQL Server 2025 — schema design, tuning, native JSON
CloudAzure — SQL, Blob Storage, App Service, Entra ID
Embedded AIModels chosen per job on best fit — Anthropic, Azure AI, and others
ToolchainBuilt with Anthropic's Claude — every generated line reviewed
StructureMicroservices with bounded contexts and clean seams
DeliveryGitHub Actions — every build versioned to its commit
TestingUnit, integration, and Playwright end-to-end
Show me your tables [the data model], and I won't usually need your flowcharts.
Frederick P. Brooks, Jr.

Our founder studied under Brooks at UNC. It's the first question we ask a founder — and the reason we start with the data model, not the screens.

§ 05

What we're like to build with.

Not slogans. These are the habits that decide whether a company can still move quickly in year three.

Order of work

Foundation before features

Speed early is fine. Speed on a wrong data model is a rewrite you'll pay for right when you can least afford it.

Alignment

We're paid when it works

Holding a stake changes the advice. We'll argue against work that bills well and builds nothing, because it costs us too.

Handover

Built to be taken over

Every build is versioned to its commit and tested where it counts, so your own team can pick it up without a hostage negotiation.

If you're building something

Tell us what you're building.

We take on very few companies a year. Send us the idea and where you are with it, and we'll tell you straight whether we're the right partner — and what we'd want to own.