Keaton Taliaferro · AI tools for nonprofits
Not chatbot wrappers — verifiable tools. Every project backs its output with something a human can confirm in seconds: a verbatim citation, a rubric you can read, a signal you can trace.
Why it matters: the easy version of "AI for nonprofits" is a thin wrapper around an agreeable model. In a domain where a wrong "yes" costs an organization a grant or ships a broken laptop to a family, agreeable-and-confident is dangerous. So I separate the model from the decision — the model observes, deterministic code you can audit decides.
Turns a funder's RFP into a checklist of every requirement, then checks your draft against all of them.
Extracts each requirement with a verbatim RFP citation, and refuses to mark anything "met" unless it can quote your own draft back — catching the model fabricating coverage, live.
A ~$21 edge sensor that predicts a failing food-bank fridge hours early — with zero AI at runtime.
Predicts failure from the warming trend, ignores door-opens so the alarm stays trusted, and alerts offline — enforced by a source-level test. Same logic ships in browser JS and ESP32 firmware.
Consistent, standards-based grading for donated computers, so a good machine isn't scrapped and a broken one isn't shipped — my signature project, applying industrial quality-inspection discipline to e-waste.
The vision model only observes; a versioned rubric with severity floors decides — and no blocking-defect device can ever be graded "refurbish," enforced against the model and a human override.
Plain language answers about SNAP and energy assistance eligibility, in English and Spanish, for people the rules were written to help and not to be readable by.
Answers only from official rule text it retrieved, quoting the rule every time. Ask about a program it hasn't loaded and it refuses and routes you to a caseworker. Any categorical "you qualify" is blocked, in both languages.
100% client-side. No servers, no telemetry, no API keys. In "live" mode the only network call is to a model on your own machine (local Ollama).
Verification is plain, inspectable code. No AI is required to check the AI — each safety engine is a few hundred readable lines with automated tests.
Runs free, forever, on hardware you own. Nothing to subscribe to; deployable by a nonprofit at $0/month.
Data-safe by design. Demos use public or synthetic data only — no client's confidential material anywhere.