DRYCODEWORKS
Daniel Young

Systems that survive production.

Teams usually bring me one hard problem, then keep me on to build the rest, often as their fractional CTO. I work across the whole system: the product and its interfaces, the data layer down to database internals, the AI models inside it, and the infrastructure down to the operating system and the network.

Daniel Young
Daniel Young · Founder, DRYCodeWorks
Previously at
AWSGoogle
Recent clients
SazabiFrost SolutionsMetopio

What I build

MCP servers that give AI assistants your company's data with each user's permissions, agent backends, and vision models trained and run in production.

Permissions outside the model
AI clientsauthenticated callersHosted MCP servercompany sign-inService accountread-onlyDatabaseread-only dataWHERE EACH REQUEST IS CHECKEDsign-inauthorizationgrantsresult capUser Aauthorized requestansweredUser Bnot authorizedrefusedUser Crequests restricted datadeniedchecked by the gateway and the database, not the model
Illustrative access boundary. A hosted MCP server authenticates the caller, checks authorization, and uses bounded read-only database access. The model cannot grant itself permission.
Learn more →

How I work

I build the AI tooling I work with: parallel agent sessions, each in its own sandbox with its own network identity, and nothing merges until the tests pass. I write about how it works, and I build the same orchestration for clients.

How an engagement runs ↓
Live, last 30 daysUpdated hourly
Commits that passed CI2,672
Agent sessions836
Changes merged1,887
Open source, MIT licensed

Tools from my workflow

I built these to run my own agents, then open-sourced them.
ClickHouse schema migrations in plain SQL: bootstrap, rollbacks, and a check that every statement is safe to run twice.
~/ch-migrateGitHub →
$pip install ch-migrate-cli

What shipped

All case studies →
Performance rescue for an IoT platformI led a three-person team rebuilding an IoT platform for peak traffic, without moving off SQL Server.
Average API response time12 s → 150 msIoT platform · In productionRead the case study →
Infrastructure for a production vision modelLLM labeling, versioned training data and a promotion gate behind a custom vision model, now run by the client's own team.
Labels wrong: frontier models → this model15–20% → ~4%IoT platform · In productionRead the case study →
Moving IoT analytics from SQL Server to ClickHouse1.3 billion sensor rows moved to ClickHouse Cloud with no downtime, for $2,200 a month less.
Dashboard query time45 s → under 2 sIoT sensor platform · In productionRead the case study →

I also write guides and essays on ClickHouse, AWS and AI coding agents.All articles →

In a client's words

“In our first season with DRYCodeWorks, we were able to increase our units sold by 300% and completely eliminate service interruptions.

Before partnering with DRYCodeWorks, we were grappling with a complex web of legacy code, technical debt, and growing instability of our web platform. Dan led our team through a complete redesign of much of our critical infrastructure that radically improved the performance and reliability of our platform.”

Mike BottMike Bott, CEO of Frost Solutions

Tell me what you're building

A free 30-minute call with me, not a sales team. Bring the diagram or the prototype. My engagements run long and personal, so I'd rather know we fit first. Either way, you'll get a Slack Connect invite for later questions.

Book a call

How an engagement runs

First callFree, 30 minutes. You tell me what's in the way, and I'll tell you whether it's a problem I'd take on. It's a fit only if we both say so.
ShapeEngagements flex to the problem: an assessment, a build, a rescue, or working inside your team. We pick the shape after the call, not before.
DeliveryEverything ships through your repos and your review process, with the tests and docs your team needs to run it without me.