ABRAR ABIR

I’m a research software engineer (RSE) on Genomic Data Science Core at Sidra Medicine, which looks like full-stack development over the anarchy of fragmented high-stakes data (part of the work made it into a poster at SARF ‘26). Genomics is a big-data problem in its own right (on par with astronomy/social media) as sequencing gets accessible, yet more data hasn’t made drug discovery any more efficient (if anything, it’s gone the other way - following Moore’s law backwards!). Part of that is a tooling problem: life science doesn’t have a sustainable ecosystem for software development, so most tools get built for temporary scope then abandoned, and/or aren’t installable in a user friendly manner.

On top of it, life sciences demands a level of transparency and reproducibility that isn’t always straightforward. Plus, most of what I touch sits upstream of critical decisions at precision medicine and/or scientific discovery. All of it makes it so much fun compared most corporate SWE gigs ever get to be (let’s not even talk about bullshit jobs).

On the side, I explore the wild landscape of open-source tools trying to shape conventional software engineering around the scope of life science. One such attempt is a desktop app for AlphaGenome I’m building with the molgenis folks at UMCG (a follow-up to my own attempt), about making predictions accessible to non-technical users. Another scratch work is about trying to make LLM agent workflows robust and reproducible, chasing the determinism that genomics demands. Also, a weekend detour was hackathon project where I built the foldback caller to detect ONT artifacts that SV callers misread as real.

Not long ago, I had a good sprint with ML research. An NLP course pulled me into shared tasks (ArabicNLP ‘24, CLEF ‘24), and one thing led to another: a QNRF-funded CMUQ × QU collaboration on automated essay scoring for Arabic produced a new dataset (EACL ‘26). Turns out SOTA methods built for Eurocentric languages doesn’t really work out-of-the-box outside the language family.

Among other things, I’m (currently) curious about epigenomics × ML and agentic genomics. Downtime is lettering, chess, or hanging out in the bio/arXiv.

As usual, I’m open to collaborating on over-ambitious projects, discussing nearly-impossible ideas, or connecting over almost-thoughtful conversations.