Senior Scientist, AI-Assisted Lab Optimization

• Division: Detection, in close partnership with Secure Bio AI
• Start date: as soon as possible
• Starting compensation: $180,000 to $210,000, depending on experience
• Work location: In-person in Cambridge, MA

Secure Bio is a 501(c)(3) non-profit dedicated to protecting civilization from catastrophic biological risks, particularly pathogens designed to evade detection. We operate two divisions, Secure Bio Detection and Secure Bio AI, and serve as a technical advisor to the US and allied governments.

Our Detection team co-leads the world’s largest wastewater metagenomic sequencing network, monitoring more than 20 million people across 30 or more sewer sheds and generating over half of all public untargeted wastewater sequencing data worldwide. We have twice flagged aberrant engineered sequences in the wild and traced them to their sources.

Our AI team builds industry-standard evaluations of frontier models’ biological capabilities, including the Virology Capabilities Test and the agentic ABC-Bench, now cited in the model cards and risk frameworks of every major AI lab and referenced in US Congressional hearings.

The two divisions operate differently. Detection is in person, lab based and growing quickly. AI is fully remote, distributed from Berkeley to Berlin to Melbourne. This is the first role to sit formally across both, and the first time the two teams have worked toward a shared goal, which is a significant part of why we are hiring at this level.

Metagenomic surveillance is the most robust defense we have against novel or engineered biological threats, but its value depends on speed and sensitivity. Against a pathogen with a three day doubling time, shaving even eleven days off the path to a confident answer lets authorities intercept an outbreak while it is roughly thirteen times smaller. The downstream impact of speed is, quite literally, exponential.

At the same time, the rapid development of general AI capabilities has raised serious concerns that bad actors could use these tools to cause harm through biology. Large language models already exceed human experts on lab oriented practical knowledge evaluations such as our own Virology Capabilities Test ( VCT). Whether that translates into meaningful uplift in a working laboratory remains unresolved, and it is one of the more consequential open questions in biosecurity today.

We are launching a proof of concept program to answer it, using our detection pipeline as the case study, and we are hiring a senior wet lab scientist to lead it. You will be embedded in an AI driven R&D loop, using frontier models as collaborative partners to optimize the wet lab protocols underpinning Secure Bio Detection, while generating the data and intuition that guide Secure Bio AI’s research into biosecurity relevant uplift.

Two things come out of this work. A detection pipeline that is measurably better than it was. And a rigorous, honest account of what AI uplift looks like for an expert biologist doing real work, including where the models mislead as well as where they help, which is the part the field is currently missing.

You will have a great deal of independence, and discretion to devote significant resources toward the success of the project. You will also serve as a key link between the two major groups within Secure Bio, which is why we place unusual weight on communication and creativity in the requirements below.

At its heart, this is an opportunity to apply cutting edge tools to an important problem, and to see the results of your creativity and dedication produce tangible, measurable benefits to biosafety.

You will lead a hands-on, AI-in-the-loop wet lab R&D program, owning it from experimental design through…