I run Abera Consulting in Toronto and build AI document systems where a wrong answer costs real money. The through-line is verification: extraction pipelines with citation gates, and eval harnesses that catch drift before users do. PermitCheck puts the exact bylaw line under every permit verdict. Enacted computes clause-level diffs of Ontario law from the source text instead of summarizing it. NoeticMap triaged 65,074 research papers down to 2,233 full extractions and won't publish a claim whose citation doesn't resolve. The citation gate behind that one became sourcecheck, a standalone library.

The conviction came from getting burned. An early pipeline of mine fabricated 4 of 5 citations in a spot-check, and the output looked completely fine. Since then I don't ship a pipeline without a deterministic check on what comes out.

I've been in software development since 2015, and I got in sideways. I was working as a tender analyst in Ethiopia, where I'm from, digging through millions of rows of import records to work out what people were paying in duties on pumps and generators, and from that, what bid prices they could be submitting. Each pass took about three days by hand. So I spent a month learning Ruby on Rails and built HS Ethiopia, a dashboard that turned those three days into a few searches.

The tool also surfaced something I wasn't looking for: by 2016, Ethiopia's imports had grown to almost five times its exports. Not long after, the government stopped publishing the data. That experience convinced me that software gives you more leverage to change things than anything else I'd touched, and I've been building, and teaching people to build, ever since.

That led to startup work in San Francisco, then senior product engineering at Ava Labs, where I worked on avacloud.io, core.app, and the wallet extension. Along the way I founded GrocerBox, an e-commerce platform for independent grocers, and sold it.

Mentorship matters to me. I mentored at Lighthouse Labs from 2019 to 2021, and teaching is still how I find the gaps in my own understanding. It's why the fundamentals series runs in public instead of in a private notebook.

I'm a father, and Toronto is home. The rest of what I believe about this work is in the writing: one essay a week, mostly about what broke and what I did about it.