AI Enablement & Governance | Workflow Automation, Evaluation & Human Review | Environmental Science & Climate Resilience
I work on practical AI and data workflows that help people make decisions with visible evidence and clear review responsibilities. At North Coast Opportunities, I coordinate outreach for the BUILD program in Lake County, including SolarCHILL. I am completing a B.S. in Environmental Science at Southern New Hampshire University.
My portfolio connects environmental data systems with bounded AI workflows and evaluation. Development includes AI assistance; prototype behavior, software test results and field validation are distinct kinds of evidence.
| Project | What it demonstrates |
|---|---|
| Clear Lake Watch · live dashboard | Environmental dashboard prototype for Clear Lake, CA, using public hydrology, state HAB reports and GIS context. The public-data snapshot was generated May 5, 2026. Freshness checks warn on stale sources; source metadata and map review cues support human review. Late prototype / early MVP; not current lake conditions or official public-health guidance. |
| community-monitoring | Intake-side counterpart: simulated field-sensor pipeline with schema validation, semantic QA/QC, station provenance checks, and human-review routing before public use. |
| framework-aligned-emissions-calculator | Local-first Python prototype for Scope 1 stationary combustion, location-based Scope 2 electricity, and educational spend-based Scope 3 Category 1 screening. Not a certified GHG inventory tool. |
| Project | What it demonstrates |
|---|---|
| TriageCore | Local-first research workbench for bounded agent workflows, authorization checks and reviewable execution records. Research and prototype behavior; no enterprise deployment or certification claimed. |
| agent-control-evals | Evaluation tooling comparing static exported control-plane decision records with separately defined expected outcomes. Tests decision-contract conformance in defined scenarios; it does not verify that reported decisions were enforced or establish an independent human audit. |
| SafeGate Edge | Verifiable approval boundary for high-risk agent actions: hash-bound human approvals, fail-closed mutation blocking, tamper-evident evidence bundles. |
| SafeTask | Hazard-first local safety evidence pipeline: fail-closed redaction, human-review gates, privacy-preserving export. |
Shared terms are defined once in the trust vocabulary so definitions don't drift between repos.
- Claims match code. I check documentation about enforcement against the implemented behavior and label proposed controls as proposed.
- Distinguish warnings from blocks. I distinguish code-enforced controls from warnings and identify decisions that require manual review.
- Boundaries before features. Privacy, governance, and non-claims are written down before intake pipelines are built.
AI enablement · workflow automation · evaluation · environmental science · GIS — Lake County, CA