AI Systems Advisory
I advise teams building agentic and LLM-powered systems: architecture, tool and orchestration design, and evaluation.
Hands-on experience
- › Agent orchestration on AWS Bedrock AgentCore
- › Model Context Protocol (MCP) server design and implementation
- › Retrieval and evaluation architecture for high-stakes, compliance-sensitive domains
- › LLM fine-tuning and reasoning-trace experimentation (RLHF, GRPO-family methods)
- › AI delivery for government and regulated-industry service workflows
How I work with teams
Architecture review
A read of an existing agentic or LLM system: orchestration boundaries, tool design, failure modes, and what breaks when it meets real users.
Advisory
Recurring time with a team that's already building. Design decisions, tradeoffs, and a second set of eyes on the parts that are hard to undo later.
Hands-on consulting
Building alongside your engineers on prototypes, evaluation harnesses, MCP servers, and retrieval pipelines.
I scope engagements per project. Most start with a short conversation about what you're building and whether I'm the right person for it.
Featured: Research
Long-form research where the claims are tested against ground truth that can be checked — benefit formulas defined in statute, and a small model whose failures are published with its scores. All research →
Weight-Space Programming for the Social Safety Net
An independent test of the claim that AI models can write, read, and edit neural network weights directly, run against US benefit formulas where the correct answer is defined by law. Written weights were exact. Trained weights failed in ways average accuracy hides. And one step of the law cannot be represented at all.
Nine Versions of a Redaction Model
I trained my first model to redact PII and PHI on-device: an 11-million-parameter token classifier that fits in 11.3 MB. Nine versions of synthetic data, strict span scoring, and a failure ledger got it to 0.974 full-value recall on my golden set, with the misses still named in the post.