Systems and tools
Systems and tools for building agents, evaluating their behavior, and organizing human oversight.
Operational workflows, software development, and reproducible experiments.
An operational workflow for agent evaluation
Scout finds conversations relevant to a project's work and drafts replies for human review. Human grades, feedback snapshots, and offline replay support experiments on relevance, classification, and policy.
What to inspect: The operational workflow, grading/replay path, and rubric study.
A workspace for agent-driven software development
Oakridge organizes software work into durable workflows, stages, artifacts, and review. Its operator workspace brings agent sessions and their outputs together so the human can inspect the work and decide what happens next.
What to inspect: Workflow definitions, artifact handoffs, and the operator workspace.
Reproducible experiments for agent systems
Assay runs paired experiments against declared plans and records the inputs, outputs, evaluations, and costs. It supports investigations into coding behavior and classification with results that can be inspected and reproduced.
What to inspect: Experiment plans, paired comparisons, and the coding-model study.
Foundations for executing agent work and organizing oversight.
A modular runtime for agent loops and pipelines
A modular runtime for agent loops and orchestrator-controlled pipelines, with tracing, grading, feedback, and interchangeable adapters.
What to inspect: The two execution models, small interfaces, and tracing/feedback path.
Task-specific autonomy earned through evidence
PAA describes how automation can earn task-specific autonomy from evidence, with explicit oversight, promotion, and demotion.
What to inspect: Task contracts, evidence requirements, and oversight transitions.
Code visualization, payments, privacy, and cryptographic developer tooling.
Interactive maps of codebase architecture
SysVista scans codebases into interactive visual maps of models, services, transports, and transforms, with source details and relationships to inspect.
What to inspect: The codebase scanner, system/flow graphs, and source inspection.
API access gating using x402 micropayments and ERC-8004 onchain reputation. Flat fees, reputation-based pricing tiers, and combined routes where reputation unlocks discounts.
Anonymous survey platform using Semaphore protocol for cryptographically verifiable anonymity. Enables organizations to collect honest feedback while mathematically guaranteeing respondent privacy.
Development toolkit for Circom circuit development and optimization.
Shell-based toolkit for Noir circuit development. Constraint profiling, benchmarking, and CI integration for production ZK workflows.
Selected engineering contributions from previous roles.