Built In-House: How Our Own AI Is Expanding What We Offer

Built In-House: How Our Own AI Is Expanding What We Offer

Every week, people serving time document their preparation on prisonprofessors.org. They write journals, complete coursework, and draft release plans. My job is to build the technology that helps that work get seen, shared, and turned into resources others can use. Here's what our team is building and why we chose to run our own AI to do it.

The profiles platform

Everything starts with the profiles platform. Participants use their profiles to record the choices they're making while they serve their sentences. Each journal entry, finished course, and revised release plan adds to a record that shows who they're becoming.

That record has value beyond the person who wrote it. Someone just beginning a sentence can learn a lot from how another person structured their days or explained their plans for life after release. We wanted to build a way to bring that work forward.

Profile Leaders

On October 23, we launched a daily series called Profile Leaders. Each morning, our system reviews profiles and selects one participant whose documented work shows a clear commitment to preparation. It drafts a short post recognizing that work. Each post draws out a lesson and ends with a challenge question for readers, and we share them on our website and social media.

The system also drafts a letter of recognition for each participant. Michael Santos reviews every post and every letter before anything goes out. When he edits the writing, those changes help the system learn the voice and standards we expect. Each new draft starts closer to what he would write himself, and Michael keeps the final say.

Turning individual work into shared resources

We built Profile Leaders to do more than recognize one person a day. The same system turns participants' documented work into lessons others can follow. We collect those lessons in a series called Prep for Success after Release.

Over time, that series will grow into a library of practical guidance built from real people's preparation, not theory. A person facing a sentence can read how others chose courses, built routines, and explained their release plans. Then they can adapt those examples to their own situation.

Why we run our own AI

The people we serve trust us with sensitive information. Profiles can include names, details about where someone is held, and personal stories about their lives and their cases. Protecting that information shapes how I build everything.

So our team runs its own AI models on computers we operate in-house. When the system reads a profile to draft a post or a letter, that information stays on our own machines instead of passing through outside AI services. Participants share their stories with us, not with a third party's servers.

Running our own models has a second benefit. It reduces our reliance on paid, per-use AI services, so we can run this work every day instead of rationing it. And the same in-house setup gives our team a place to build and test new tools before they reach anyone. We can try an idea, find where it falls short, and fix it before people depend on it.

What's next

We're expanding the resources available through the profiles platform. We're also building tools to support reentry, so participants can carry the work they documented inside into their lives after release. We'll share those tools when they're ready, not before.

Technology can't do the work for anyone. No system will write a person's release plan or finish their coursework. What it can do is help us recognize that work faster, share it more widely, and protect the people who trust us with their stories. That's what we're building toward.

If you or someone you love is serving time, start a profile on prisonprofessors.org and begin documenting your preparation. Your work could become the example that helps someone else get ready for home.