At Lawrence Berkeley National Laboratory, a quantum materials pipeline built around a robotic arm shuttling wafers through a sealed cleanroom is now screening bacteriophage candidates against drug resistant urinary tract infections.
Aeron Tynes Hammack spent the last few years teaching a robotic arm to build quantum computers. The arm shuttles an 8-inch wafer through a sealed, robot-only cleanroom at Lawrence Berkeley National Laboratory's Molecular Foundry, depositing, etching, and inspecting materials in a single pass. The point was Josephson junctions for qubits. The pipeline he built to make that work is now screening viruses that kill antibiotic-resistant bacteria.
Hammack, a physicist by training and the Molecular Foundry's Interim Facility Director of the Nanofabrication Facility, described the redirect in a Berkeley Lab News Center Q&A published June 9. The shift cuts across the customary boundary between physics instrumentation and clinical microbiology, and it rests on a simple structural bet: that the high-throughput, closed-loop, robot-driven experimental pipelines designed for materials discovery can be repurposed for biological discovery at the same scale.
The bet has a real commercial and clinical stake. Hammack co-founded EpiBiome with Nick Conley; the company was acquired by Locus Biosciences, which is running clinical trials on a bacteriophage cocktail for urinary tract infections caused by uropathogenic E. coli. According to Hammack's account in the Q&A, Locus has tested a six-phage cocktail, a mix of wild-type and engineered phages, against 356 UPEC strains in the lab and reported 96.4% effectiveness, with 29% of the strains being multi-drug resistant. Those figures are Hammack's own summary of the company's results; they have not been independently verified within the article's source packet and should be cross-checked against the underlying Locus disclosures or peer-reviewed paper before publication.
The pipeline that produced them started in quantum materials. Berkeley Lab researchers, including Hammack, used the Molecular Foundry's quantum information science (QIS) cluster tool to fabricate Josephson junctions out of hafnium, a superconductor-insulator-superconductor stack that is one of the candidate recipes for stable qubits. An arXiv preprint (2510.25203) documents the work; arXiv preprints are not peer-reviewed, so the result should be cited as preliminary. The QIS cluster tool itself, the same piece of hardware now being redirected at phage research, is described in a separate Berkeley Lab News Center article as a closed-vacuum cleanroom in which a robotic arm moves wafers between deposition, etch, and analysis stations, automating materials research that was previously hand-pipetted.
The redirect to biology was not metaphorical. Hammack's team used automated liquid-handling robotics, computer-vision colony counting, and optical-density assays to screen 2.5 million phage-host combinations, a workload that would have been unmanageable by hand. The method is described in a Nature Communications paper linked in the Q&A. The scale matters: traditional phage hunting is artisanal, with researchers testing a handful of candidates against a few bacterial strains. Screening 2.5 million combinations changes the search from a craft to a dataset, and the dataset is what makes it possible to pick a six-phage cocktail that survives contact with hundreds of clinical isolates.
The historical irony is hard to miss. Bacteriophages, viruses that infect and kill bacteria, were discovered in 1915 and 1917, before Alexander Fleming's penicillin became widely available in the 1940s. Phage therapy was largely sidelined in Western medicine once broad-spectrum antibiotics arrived, and it has spent decades as a curiosity while bacterial resistance to those same antibiotics has grown into a serious public-health problem. The phage work now coming out of the Molecular Foundry is not the first attempt to revive it, but it is the first time this particular pipeline, a closed-vacuum, robot-driven, AI-assisted materials-discovery rig, has been used as the engine.
Hammack credits his postdoctoral work at the Molecular Foundry, where he was exposed to combinatorial chemistry and liquid-handling robotics for nanoparticle synthesis, with giving him the tools to move phage research from artisanal to high-throughput. That origin is the structural insight the piece turns on: the same laboratory infrastructure, originally aimed at quantum materials, can be re-aimed at a clinical problem with little change to the underlying automation.
The wider ambition is institutional. The Phage Foundry, a multi-institutional program led by Berkeley Lab, is building a public phage biobank with receptor-usage, gene-essentiality, and AI/ML effectiveness data, with Vivek Mutalik named on the team. The Phage Foundry is pitched as a countermeasure platform, a place where academic and government labs can pool phage candidates, host-range data, and machine-learning models the way the National Institutes of Health's genomic databases pooled sequence data in the 1990s.
What to watch: whether the Locus Biosciences cocktail survives Phase 2 and Phase 3 trials; whether the 96.4% in-vitro figure holds up in human patients, where pharmacokinetics, immune response, and bacterial ecology are not modeled by an optical-density assay; and whether the closed-vacuum cluster tool pattern, one robot, one cleanroom, many stations, becomes a template that other DOE Office of Science user facilities adopt for other cross-domain discovery problems. Berkeley Lab, a multiprogram national laboratory managed by the University of California, was founded in 1931 and lists 17 Nobel Prizes across its history. The cross-disciplinary bet described here is, for now, a single physicist's pipeline and a single biotech's trial. The structural question is whether the pattern outlasts either.