Founded 2024, Baltimore MD
We built Genolux because phage display is broken.
Dr. Elena Voss spent years leading computational design at a clinical-stage biologics company — watching $800K–$2M campaigns run 6 weeks of experimental screening on targets where AlphaFold2 had already made interface prediction tractable. In early 2024, she and Marcus Chen started Genolux to close that gap: structure prediction existed; nobody had built a pipeline that made it usable for a discovery team's actual workflow.
Mission
Make therapeutic antibody design fast enough that small teams can actually iterate.
01
Interpretable scores
No black box. Every ΔΔG prediction traces to a Rosetta energy term. Every developability flag has an explicit computational basis. Scientists can read the decomposition and judge for themselves.
02
Scientific honesty
We publish our benchmarks, including where we fail. CDR H3 loop prediction degrades beyond 14 residues. We say this on our Science page. Hiding limitations is how the industry built a credibility deficit with computationalists.
03
Discovery-team focus
We are not selling to enterprise procurement. The Explorer free tier and $490/month Discovery plan are priced for individual antibody engineers, not procurement departments. That's a deliberate design choice.
Location
Baltimore and the JHU computational biology ecosystem
Genolux is headquartered at 100 East Pratt Street, Suite 2400, Baltimore, MD 21202 — within a few miles of Johns Hopkins University's Biophysics and Biomedical Engineering programs and the University of Maryland School of Medicine. That proximity is deliberate: we hire from those programs and draw on their computational structural biology and bioinformatics communities.
The Baltimore–Washington corridor has one of the densest concentrations of computational biologists outside the Bay Area and Boston — between JHU, University of Maryland, and NIH. It shapes every hire and every validation partnership we form.
We are computation-only by design. Genolux does not run plates or express antibodies. We narrow the candidate space so that when your wet-lab runs SPR or octet binding experiments, it is working a 20-sequence shortlist — not a 500-compound hit pool.
Funding
Angel-backed, focused on prediction engine and benchmark validation
$2M angel round closed September 2025. Capital is allocated to compute infrastructure, scientific hires, and the open benchmark program — because our accuracy claims only hold if we can show you the numbers and the code that produce them.
Founded 2024. Angel-backed. No VC board has an opinion about your sequence data.
Join us or work with us.
We're a small team in Baltimore. We publish openly and build the tools we wished we had.