← Selected work Research / Decision Pipeline
SignalOps — Engineering Evidence to Action Implemented a policy-driven workbench that preserves observed evidence separately from interpretation, ranks relevance deterministically, enforces channel permissions, stores durable state, and produces qualified next actions.
Public technical artifact Problem
What this was built against. Engineering and market signals are noisy; without provenance, dedupe, policy gates, and state, research becomes a pile of posts rather than an executable decision system.
System
How the work closes the loop. Channel-native evidence → normalize/provenance → deterministic relevance → permission gate → durable state → ranked action / handoff / CRM-ready export.
Evidence now
What an employer can safely inspect. Public repository documents provenance, idempotency, human-permission controls, SQLite state, ranked actions and CRM export boundaries. Recorded claim registry reports 11 unit tests + smoke + compile passed offline for SignalOps. The broader proof corpus records issue-native GitHub contribution work derived from real engineering problems. Claim boundary
What this does not prove. No claim of live CRM integration or autonomous outreach. No claim that the system scrapes or infers private data. External contribution counts should be linked to the actual contribution ledger before using a precise number in applications. Next evidence event
What would upgrade the proof. Attach a public contribution ledger and one end-to-end example from source issue → scored signal → artifact/contribution → maintainer response.
P2 — reproducible offline system; external contribution layer separate · Canonical case-study page v1