Neon — Skunkworks at Speed
We build sharp, neon‑clean systems — mapping, retrieval, testing, and agents — then ship playbooks others can reuse. Short, focused, verifiable.
Neon is our internal lab: we prototype, extract patterns, and turn them into tools and write‑ups the rest of the company can adopt without ceremony.
Projects
Project Iris
Safe data transformation.
Iris is a DSL‑driven mapping engine for structured data. It defines how JSON transforms from one schema to another — safely, predictably, and verifiably. Mappings are LLM‑generatable, validated, stored, and executed at runtime without redeploys.
Project Eden
Structured retrieval for structured data.
Eden combines vector retrieval with structured querying. It builds explicit schemas for documents and tables and supports embeddings, direct JSONB filters, and hybrid pipelines that merge both.
Project Rhea
End-to-end API verification.
Rhea is a DSL-based test runner for APIs. Suites are JSON that describe endpoints, requests, assertions, and expected behavior — authored by humans or generated by LLMs. It runs automatically, logs metrics, and versions results.
Project Adam
Structured autonomy.
Adam is a framework for building structured agent systems. It manages prompts, behaviors, memory, and policy enforcement. Agents reason, execute, and coordinate under defined rules, not loose text chains. Every decision is observable and traceable.
