O&A Consulting, LLC

Intelligence, crafted.

A boutique AI engineering studio for high-stakes domains — financial services first. We build generative AI systems, knowledge graphs, and agentic workflows with the polish of luxury hospitality — because how software feels is part of whether it works.

10+years engineering AI & software
400K+documents graphed in production
Hours minutesresearch, citation-backed
Scroll
01 The studio

Engineering depth. Hospitality-grade experience.

Most AI consultancies sell headcount. We sell craft — senior, principal-level work on every engagement, from the retrieval architecture to the last pixel of the interface.

O&A pairs a decade of production AI engineering — GraphRAG platforms for one of the world's largest asset managers, fraud analytics for major financial institutions, agentic research systems — with a service philosophy drawn from luxury hospitality: anticipate, refine, and never ship anything you wouldn't sign.

We stay small on purpose. You work directly with the principal engineer, not a bench.

/ 01

Generative AI & GraphRAG

Knowledge graph pipelines, hybrid vector + graph retrieval, and citation-backed reasoning over your hardest data.

/ 02

Agentic Systems

Multi-agent workflows with validator agents that fact-check every claim before a human ever sees it.

/ 03

Trust & Evaluation

Evidence-backed, contract-driven evaluation that answers the question every agent project faces: can we ship this?

/ 04

UX Engineering

Frontend craft from a team that believes an AI system is only as trustworthy as it is understandable.

02 Selected work

Products we've shipped.

Aurum
assets/aurum-promo.mp4
Product · Consumer AI

Aurum

An AI-guided scripture study companion, built on a knowledge graph of the biblical text. Aurum doesn't just answer — it traverses centuries of interconnected people, places, and passages, and shows its sources every step of the way.

Knowledge Graph RAG, in the hands of everyday readers.
Knowledge Graph RAGGenerative AIProduct Design
Read the case study
trustkit
assets/trustkit-promo.mp4
Product · AI Assurance

agent-trustkit

The definition of done your agent project is missing. Stakeholders co-author a task contract; the evaluator collects quoted evidence — never black-box grades — and the report states exactly what the agent is approved for, and what it isn't.

98% accurate is not the same as trustworthy. Trust is a decision — and decisions need evidence.
Agent EvaluationOpenTelemetryBedrock AgentsCI Trust Gates
Read the case study
Case Study
assets/pipeline-platform.png
Case Study · Enterprise (anonymized)

The delivery team that ships itself

A multi-agent platform that takes a feature from written requirement to reviewed, tested, integration-ready code — automating a whole chain of dev and product responsibilities, with human-in-the-loop gates at every decision that matters.

Agents execute. The harness enforces. Humans keep the signature.
Multi-AgentOrchestrationHITL GatesClaude Code
Read the case study
Taming the Agent
assets/taming-hero.png
Method · Engineering Playbook

Taming the Agent

How to build with AI coding tools — Claude Code, Copilot, Codex — without inheriting their chaos. Sixteen classic engineering guardrails, from strict types and OpenAPI contracts to security scans and automated publishing, enforced automatically against the agent's workflow. Proven daily in production on Aurum.

Speed without structure is technical debt on autopilot.
AI-Assisted DevTypeScriptSecurity ScansCI/CD
Read the case study
03 Active research

We do the science, not just the demos.

Metrology · LLM Agents

Can you trust the tools that tell you why your agent failed?

The field has produced many failure-attribution methods — and almost no measurement science. Our attribution-audit study asks the questions nobody has checked:

  • Do published attribution methods even agree with each other?
  • Does the same method, re-run, name the same culprit?
  • Do imagined counterfactuals match actually-replayed ones?
  • On traces with planted defects, what is true accuracy?
Knowledge Graph ETL

Unstructured text, into queryable intelligence — automatically.

Ongoing work on automated knowledge-graph construction: pipelines that turn hundreds of thousands of documents into semantically connected, queryable graphs.

  • 400,000+ newswire articles graphed in one production pipeline
  • Entity extraction, clustering, and community structure (Louvain)
  • Relationships derived from cluster, entity, and metadata signals
  • Built for agentic retrieval — graphs an LLM can actually research
04 Where we focus

Built for domains where being wrong is expensive.

Financial services is home turf: GraphRAG research platforms for one of the world's largest asset managers, fraud and AML analytics for major financial institutions, and agentic reporting systems built to survive risk and compliance review. We bring the same rigor to federal primes as a specialized subcontractor — generative AI, knowledge graph engineering, and agent trust evaluation for the AI-governance questions agencies are now required to answer. Capability statement available on request.

Asset Management Fraud & AML Federal · NAICS 541511 AI Trust & Assurance
05 Contact

Let's build something
worth signing.

info@oandaconsult.com
Charlotte, North Carolina +1 (786) 226-3374 Email Aurum