AI & real-time data systems

Software that survives contact with reality.

Anyone can stand up a demo. The value is the load-bearing 10%: the data pipeline, the failure paths, the thing still up under real traffic. That's where we work, backed by thirty years building the systems businesses bet on.

a production-software studio

30+
Years building production distributed systems
10s of K/s
Events/sec through a real-time Flink aggregation engine
200+
Tests across shipped open-source tools
8
Published engineering titles (Wrox), editorial board

Demos are easy. Production is the job.

The hard part was never the demo. It's the system around it: the unhappy path, authorization on every route that emits data, idempotency, observability, the pipeline that feeds it. That's ordinary production engineering discipline, three decades of it, so what you ship is something you can actually run, not a prototype that quietly breaks at 3 a.m.

// What we do

From idea to a product you can run, in weeks.

01 / Build

Production SaaS MVP

Your idea (or a stalled prototype) taken to a deployed, customer-ready product: frontend, API, data, auth, billing, and infrastructure. Fixed scope, fixed timeline, shipped in weeks, not quarters.

02 / Specialty

Data & AI Products

The hard builds, where the data or the model is the product: ingestion and real-time pipelines, geospatial and search, RAG and agents, built to stay fast and accurate once real users arrive. (See ForeclosureBase.)

03 / Harden

Production Rescue

A demo that won't survive real users? We add the load-bearing 10%, authorization, idempotency, observability, CI, and a real deploy, so what you have becomes something you can actually run.

// Built, shipped, run

Selected work.

ForeclosureBase Production proptech SaaS · solo build
A Zillow-style platform for distressed-property investors: a multi-source county-records pipeline, sub-100ms PostGIS map search, Stripe billing, and server-side tier-gating, shipped to Kubernetes with GitOps, CI, and full observability, designed, built, and run single-handed.
~$500k scope · solo
Real-Time Event Aggregation Engine High-throughput streaming · Flink
A 48-operator Flink DAG with 13+ keyed state descriptors, aggregating a high-volume event stream into running totals that drive live downstream state, in real time.
10s of K events/sec
StreamOps Agent AI ops agent · open source
An AI operations agent for streaming infrastructure built on 13 enterprise patterns, with a 234-test suite. Public, v0.1.0.
234 tests
Codebase RAG Retrieval over code · Claude
A RAG CLI with AST-aware chunking, ChromaDB vector storage, and multi-turn chat over a codebase using the Claude API.
AST chunking
Flink State Inspector State tooling · open source
A tool that reads, analyzes, and diffs keyed and broadcast state from production savepoints, locally, no cluster needed. React UI, 4 backends.
210 tests
GCS Checkpoint OOM Deep diagnosis · Flink
Traced a checkpoint crash through four bugs across Google's client library, the Hadoop GCS connector, and Flink's state uploader, each leaking 67MB buffers.
4 bugs, 1 root cause
// What's behind it

Engineering depth, not a slide deck.

Streamwright is a production-software studio. Behind it is three decades of building the distributed and real-time systems businesses bet on, across healthcare, retail, finance, and high-throughput real-time data. Apache Kafka and Flink, big data, regulatory pipelines, the systems that can't quietly go down.

That foundation includes years as a published author and editorial-board member at Wrox in the early .NET era; we still treat clear explanation as part of the craft. Today that same rigor goes across the stack: AI and RAG systems, the real-time data infrastructure underneath them, and complete products shipped end to end.

If you have a system that needs to actually work, not just demo well, that's exactly the kind of problem we take on.

// On the name

wright (n.): a maker; a skilled builder. As in shipwright, millwright, playwright. Streamwright is a maker of streaming systems: that same craft tradition, applied to real-time data and the AI built on top of it.

// Let's talk

Have a system that needs to actually work?

Tell us what you're building. You'll get a straight answer on whether and how AI fits, and what it would take to ship it for real.