Archos Labs

The data team youdon’t have yet.

Most startups and small businesses don’t have a senior data person. They have a founder making data decisions, a developer who set up the database, and a growing problem nobody has named yet. I step into that gap.

Rob Angeles, photographed against the Melbourne skyline at dusk.

Principal · Archos Labs

Rob Angeles

25 years across healthcare, financial services, government, retail, and technology. I've been the data architect on programs with eight-figure budgets and the sole architect at early-stage startups building from scratch. The problems are the same at every scale — data nobody trusts, systems nobody has documented, and decisions being made without a foundation underneath them.

I work with a small number of teams at a time. Startups and SMBs who need a senior data person without the full-time salary, or organisations with a specific data problem that needs to be scoped, fixed, and handed back.

Same person on the call. Same person doing the work. No handoff. No junior team behind the pitch.

  • CDMP
  • Kimball
  • Data Vault
  • Databricks
  • Snowflake
  • dbt
  • Anthropic
  • Data Architecture
  • Data Governance
  • Sovereign AI
  • AI Agent Development
Find me

Work I’ve delivered.

Six engagements across 25 years. From early-stage startups to large programs. Anonymised by request. Specifics on the call.

Startup · Sole architect

Data architecture from scratch.

Joined an early-stage platform as sole data architect. Built the transactional model, analytics layer, and full migration from the legacy system. No team. No handoff. Four years as their data person.

4 years · greenfield to production
SMB · Reporting

50 reports. No analytics layer. Fixed.

Operational and reporting queries on the same database. 50+ manual reports with no self-service. Built the OLAP layer, separated the concerns, migrated to Power BI.

50+ reports · self-service from day one
Tech startup · AI readiness

Data wasn't ready for the model. Made it ready.

Multiple disconnected systems. No unified view. An ML model waiting on data that wasn't clean. Built the integration layer and the ML-ready data model.

Single unified view · ML pipeline unblocked
Enterprise · AI agent

COBOL lineage in 3 months.

Built an AI agent that reads COBOL source code, extracts business rules, and outputs a clean data model ready for implementation. Team estimate was 6–9 months.

3 months · estimate was 6–9
Enterprise · Platform migration

12-month migration delivered in 6.

Data foundation work done upfront. Business-rule fidelity preserved end-to-end. The platform work ran clean because the data layer was right before build started.

6 months · half the original timeline
Solo build · Sovereign AI

Fine-tuned model. Offline. $300 phone.

Designed and shipped a fine-tuned AI model running fully offline on budget Android hardware. No cloud dependency. No server-side processing. The consensus was it couldn't be done.

7 days · scoping to shipped

How I work.

  1. 01

    Every engagement starts with the diagnostic.

    8 minutes. No login. It tells you what's broken and whether I can fix it before any commitment is made.

  2. 02

    Scoped and fixed before work begins.

    You see the number before we start. It doesn't move. No retainer. No open-ended billing.

  3. 03

    Same person throughout.

    The person on the call runs the assessment, does the architecture work, and delivers. No handoff.

  4. 04

    Small number of engagements at a time.

    Not capacity. The work requires full attention and I won't give a team less than that.

What I believe.

The model was never the constraint. Every AI project that stalls hits the same wall — the data wasn’t ready. The governance existed on paper. Nobody could trace where the numbers came from. That work happens before the model arrives, not after it fails.

A team that hears the hard answer in week two can fix it. A team that hears it at launch cannot.

I built AI products as a solo founder. I’ve been the architect on programs that ran for years. The difference in scale doesn’t change what matters: clean data, documented rules, and someone who will tell you the truth before you spend more money.

Start with the assessment.

8 minutes. No login. If the score says we should talk, I’ll invite you to a 30-minute call. No pitch. No deck. Just a direct conversation about your data.

Take the assessment8 min · no login required
Book a call30 min · I'll tell you if it's not a fit
Take the assessmentBook a call