Change your trajectory

Data problems where information is scattered, unstructured, or trapped in formats that don't talk to each other. Think: extracting structured facts from thousands of documents, connecting data across systems, or building pipelines that turn messy inputs into something reliable and searchable.

I use LLMs where they genuinely help—extraction, matching, classification—but they're usually one piece of a larger system. If your problem is better solved with a spreadsheet or a well-written SQL query, I'll tell you that.

More at the start, less over time. Early on I need access to the people who understand the problem—what's actually painful, what the data looks like, what "good enough" means. That might be a few hours in the first week or two.

During prototyping I'll share work frequently and need feedback. Once we're building for real, involvement drops to occasional check-ins and testing. By handover, the goal is that your team understands what's running and can operate it without me.

It depends entirely on the problem. A small integration might take a few weeks; a complex data pipeline with verification workflows takes months and evolves as we learn what actually works.

Rather than give you made-up estimates, I'd point you to the PoliLoom devlog—it shows how a real project unfolded, including the dead ends and course corrections. That's more honest than a tidy timeline.

What I can promise: I ship early and often. You'll see working pieces within the first few weeks, not a big reveal after months of silence.

You do. Everything I build for you is yours—code, configurations, documentation. I prefer to build things that could be open-sourced if you wanted, and I'll actively suggest it when it makes sense. No vendor lock-in, no proprietary dependencies that tie you to me.

Both. I design and build the full system—data pipelines, APIs, and the interface people actually use. A clear UI isn't optional; it's what makes the difference between a tool that gets used and one that gets abandoned.

A conversation about the actual problem—not a polished pitch, just what's frustrating and why it matters. I work best with organizations doing something meaningful: journalism, accountability, public interest, open data, or businesses that genuinely care about doing good work rather than just scaling revenue.

If your goal is "add AI to make investors happy," we're probably not a match. If you're trying to solve a real problem and want to understand what you're building, let's talk.