AI that compounds.

Proven at Enterprise Scale

Through my work at A.Team, I've helped some of the world's largest companies put compounding AI to work.

That includes teams at Unilever, Google, Microsoft, and Coca-Cola, among others - organizations where the hard part isn't the model, it's the data, the domain, and the discipline to ship something that keeps paying off.

I bring that same playbook to your business.

What "Compounding" Actually Means

Most AI projects plateau. Compounding AI keeps getting better.

The difference is what the system is built on:

• It learns from your data - not just what's on the public internet
• It captures the insider knowledge that lives in your people's heads
• Every interaction, correction, and decision makes it more accurate
• The advantage grows over time - and can't be bought off the shelf

Generic AI is a commodity. AI trained on what only you know is a moat.

What This Looks Like

Private data plus domain expertise, wired into real systems.

I build things like:

• Knowledge systems grounded in your documents, tickets, and history
• Retrieval and RAG pipelines over private, governed data
• Expert copilots that encode how your best people work
• Agents that act inside your internal tools and workflows
• Feedback loops that turn everyday usage into a smarter model

Example Use Cases

• Answer any question about your business from your own data
• Turn decades of institutional knowledge into a searchable expert
• Onboard new hires with an AI that knows how you operate
• Surface insights buried across reports, contracts, and filings
• Standardize decisions across teams, regions, and markets
• Automate specialized work that used to require a domain expert
• Keep sensitive data in-house while still shipping real AI

How I Approach It

The moat is in the details most teams skip.

I focus on:

• Data foundations - clean, governed, and connected to the model
• Security, privacy, and compliance for sensitive enterprise data
• Evaluation and iteration (LLM-as-judge, testing, human feedback)
• Feedback loops so the system improves the more it's used
• Reliability and fallback mechanisms for production workloads

The goal is simple: build AI that keeps compounding long after I'm gone.

Tech Stack

Depending on the use case:

• Anthropic / Claude APIs
• OpenAI / GPT APIs
• Google Gemini
• Vector databases, embeddings, and retrieval (RAG)
• Python / Node.js / Elixir / Rails
• Workflow systems and background jobs
• Evaluation and observability tooling
• Secure integrations with your internal data and tools

Why Work With Me

• Hands-on experience delivering AI for enterprise teams via A.Team
• Worked with the likes of Unilever, Google, Microsoft, and Coca-Cola
• Close to 20 years building real, production systems
• Deep focus on private data, retrieval, and evaluation
• Focus on lasting advantage, not hype or one-off demos

When This Is a Good Fit

• You sit on valuable private data or hard-won domain knowledge
• You want AI that's yours - not the same tool your competitors use
• You have compliance and data-privacy requirements to respect
• You've tried AI but it never became a durable advantage
• You want something that compounds, not a one-time project