Chandler Klose / Data & AI engineering

Agentic systems.
Accountable
numbers.

I build the data platforms, agent workflows, and reporting systems behind decisions that matter.

Explore the latest work Meet Chandler

19 years leading SOX-controlled subscriber-metrics work at SiriusXM.

From source systems to the reported number.

Selected work

Open the hood.

GitHub profile

Featured / Metrics & reporting

tokenledger

One activity stream.
Every reported number reproducible.

A reference implementation for AI businesses that meter tokens and seats. A late invoice changes a metric; a bridge explains the difference; the original receipt still reproduces the earlier result.

ActivitySchema v2DuckDBBigQuery
Explore the approach

Public reference release in preparation.

One number · Three questions

  1. 01

    What did we report?

    The result, its definition, and the inputs known at that point.

  2. 02

    What changed?

    New evidence, a changed definition, and an explicit bridge between results.

  3. 03

    Can we reproduce it?

    The earlier result remains independently checkable after the change.

Data tools

AI Warehouse

A SQL workspace with a database object explorer, an AI assistant, and persistent query and chat history. Connect the conversation to the data being examined.

Inspect the repository

Agent infrastructure

GCP Fresh MCP

An MCP server connecting agents to Google Cloud tools, including BigQuery jobs, storage, and Cloud Run. Explore the implementation and its setup requirements.

Inspect the repository

Developer workflow

Aider RAG extension

Document chunking and retrieval for coding-agent context. A focused experiment in giving the agent relevant material when it needs it.

Inspect the repository

How I work

The whole stack.
The whole responsibility.

The architecture, the implementation, and the operating model need to hold together.

01 / Foundation

Data & cloud architecture

Source contracts, event models, warehouse execution, and observability. Make the path from input to output understandable.

02 / Execution

Agentic engineering

Tools, context, harnesses, and evaluation. Give agents bounded work, testable outcomes, and evidence of what happened.

03 / Accountability

Metrics & reporting

Versioned definitions, reconciliation, accountable owners, and reproducible evidence. Design for the next question and the next change.

The person behind the work

Chandler Klose.

Data executive.
Hands-on builder.

I ran the SOX-controlled subscriber-metrics work at SiriusXM for 19 years. That experience shapes how I build today: every important number needs a definition, an owner, and a trail back to its inputs.

I work across data modelling, cloud infrastructure, application code, and agent workflows. I care about how a system behaves when the inputs arrive late, the business changes, or somebody asks a question the original design did not anticipate.

CK Rock AI is where I share my independent engineering work and explore those questions in code.

See what I’m building

Start a conversation

Bring a real
data problem.

A difficult metric. An agent workflow. A platform that needs to grow.
That’s a useful place to start.

Connect with Chandler CK Rock AI company page Browse the code