Brenton Hood

Specializing in building AI systems that connect to data and processes to get meaningful work done.

As frontier LLMs get stronger, don't let your implementation be the weak link.

Selected Work

Practical examples of AI integrations.

Grounded Agentic Retrieval

Digital Librarian

A small, self-hosted LLM that answers only from a live database, not from its own memory. It features a real agentic framework: it understands user intent, picks its own tools, runs queries against an 80M-record catalog, reasons against the results, and continues until it accomplishes its task. All while showing each step of the process.

Useful for answering questions accurately from your own data, with no hallucination.

Go · PostgreSQL · small language model · visible step-by-step agentic tool loop

Credentialed users access here

Semantic Memory & Recall

CRM Memory

Shows how to give an AI a durable memory: every interaction, preference, and signal is embedded into vectors and stored in a database, so the assistant recalls the right context by meaning, not keywords — for any client, on demand.

Useful for an assistant that remembers and recalls context over time.

Next.js · pgvector · embeddings · hybrid retrieval

Credentialed users access here

Private showcase access is available for resume reviewers and hiring teams.

Ground the model

Safely connect it to source systems and data so answers come from facts, not hallucinations.

Give it tools

Let the LLM drive the process by giving it purpose-built actions it can choose from.

Make it observable

Observation builds trust: log the loop, review the decisions, and make failures diagnosable.

Adapt and adjust

Use automated review cycles to flag improvements for humans, or let the system self-diagnose and iterate when proven safe.