Building data platforms that agents can retrieve from and humans can trust.
Grounded. Cited. Governed.
Sixteen years as a data engineer. I turn petabyte-scale platforms into retrieval surfaces for AI agents, where every answer traces back to a source. I build, ship products at Axiomic AI, and write, speak, and teach about how it is done.
- 16+ years
data and AI platforms - Petabyte scale
Snowflake, dbt, Airflow - Keynote, IEEE AIC 2026
dbt Summit 2026 speaker
Building now
- contextctl: auditable, open-source context infrastructure for AI agents, with Suhas Jangoanpaper under review, IEEE BigData 2026
- Axiomic AI: three live products on one open-source agent harnessexperimental enterprise
- Built from Bits learning platformunder construction
Selected work
Systems that run in production or are on their way there. Each case study explains the problem, the choices, and the numbers.
Skills-driven agentic data engineering
productionA governed library of AI skills and sub-agents that automates the analytics-engineering lifecycle across multi-repo dbt and ingestion workflows, with humans in the loop.
Autonomous incident response agent
productionZero-touch detection, root-cause analysis, and auto-remediation of data pipeline failures using the Claude Agent SDK with Snowflake, dbt, and GitHub MCP servers.
contextctl: auditable context infrastructure for AI agents
researchA vision paper and open-source prototype arguing that agent context should be governed data infrastructure: open files for truth, Git for time, a hash-named evidence folder, one SQLite sidecar for speed, and a token-budgeted serving protocol. Under review at IEEE BigData 2026.
Axiomic AI agent harness
open sourceThe multi-agent repository template every Axiomic AI product is built from: agent prompts, MCP server configs, CI, and deploy wiring. Fork, describe the domain, ship.
Axiomic AI
An experimental enterprise: one person, one agent harness, a portfolio of shipped products.
Axiomic AI is where I test what a company looks like when agents do most of the building. Every product runs on the same harness: a multi-agent repository template with agent prompts, MCP server configs, CI, and deployment wiring. New product means fork the harness, describe the domain, ship. Three arms: Labs (product builds), Solutions (reference architectures and playbooks), Academy (tech education).
- PersonasFlowpersonasflow.com
- EstateVisionestatevision.io
- WealthPilotpilotmywealth.com
- Agent harnessopen source
Speaking
Talks on agentic data engineering and AI-ready data platforms.
- Session, Oct 13–15, 2026upcoming
Making a Data Platform AI-Ready
Applied AI Summit
- Session, Sep 15–18, 2026delivered
AI-Powered Data Development: How Agentic SDLC & dbt-MCP Transformed Our Data Engineering Workflow
dbt Summit 2026, Las Vegas
- Keynote, Aug 29, 2026delivered
Unlocking Intelligence through Data
IEEE 5th World Conference on Applied Intelligence and Computing (AIC 2026)
Writing
Practical notes on MCP, agents, Claude Code, and the data platforms underneath them.
Teaching
Two YouTube channels and a learning platform in the works.
Hands-on data and AI tutorials: Claude Code, MCP, agents, dbt, Snowflake. Also on Substack and Medium.
Conversations with people who ship data and AI systems for a living.
A structured learning platform for AI engineering fundamentals and applied builds, growing out of the Built from Bits channel. Course tracks on MCP, agents, Claude Code workflows, and AI-ready data platforms are being built now.