How to Scale Apprenticeship

Science is an apprenticeship business: see one, do one, teach one. It’s a powerful - and perhaps the only - working model to successfully facilitate the transfer of tacit knowledge. The problem is that apprenticeship doesn’t scale. A professor can only work with so many students at once before their attention gets split to an ineffective level.

AI has long been heralded as a powerful tool for personalized, hands-on learning. But AI can only learn from what it sees and when it comes to the physical execution of science. Unfortunately, we’ve never had a good way to teach AI how to “do” science.

In addition to our work in AI/robotics, tech transfer, and process improvement, Transfyr is bridging this divide, leveraging our physical AI platform to create a real-world training environment where students can receive personalized feedback to accelerate their learning of key techniques and protocols and where teachers have deep evidence to draw from when providing coaching or conducting an evaluation.

The Work of Science

In the conversation about “AI for science,” we risk placing too much emphasis on Science (with a capital S) while often ignoring the scientific work behind it.

The reality is that a skilled scientific workforce is a key driver of economic competitiveness. As large pharmaceutical companies seek to expand or re-shore manufacturing capacity, the conversation they are having with state governors has less to do with tax breaks and other monetary incentives than whether the state has enough skilled workers to staff the new facility. As just one example, since Eli Lilly commenced trials for its blockbuster GLP-1 agonist in 2020, it has announced 8 new US-based manufacturing facilities at an investment of over $43B and creating over 5,000 skilled new jobs. Partnerships with workforce development organizations were mentioned in every. single. announcement.

As such, training capacity has become a core part of states’ economic infrastructure and for these highly-skilled, apprenticeship based roles, finding a way to overcome the scaling limitations of apprenticeship is critical. Even if a state can find enough people interested in building expertise in the relevant area, the precise skills needed are often hard to train, becoming increasingly complex as new advances and modalities come online. Human error at the bench (e.g., contamination, pipetting inaccuracies, missed SOP steps) doesn't just cost time, it leads to ruined batches, expensive re-validations, and potential regulatory audit findings. Further, when an expert operator leaves or retires, they take their hard-earned scientific taste and institutional and tacit knowledge with them. Transfyr is working to solve these problems.

BioBuilder Collaboration

The BioBuilder Educational Foundation, a nonprofit focused on preparing students for meaningful participation in the biotech workforce, and Transfyr are collaborating to build a practical model for scaling hands-on biotechnology education. By deploying Transfyr’s sensor stack in the training facility, we aim to build an environment in which learners can receive personalized feedback on the skills they need to demonstrate for these highly skilled scientific roles.

Students can “review the tape” just as athletes might in a post-game film room and learn where they had a contamination event, how to recognize the different “stops” on a pipette, get a reminder to tare the scale, or even receive coaching on when they may want to escalate a question to an instructor. This novel concept recently caught the attention of the Massachusetts Life Sciences Center (MLSC), who have awarded a grant of almost $1M to support BioBuilder and Transfyr’s partnership as part of their statewide “Gamechanger” workforce-development investment.

For students, this technology allows them to gain much more practice in a safe, hands-on, personalized environment before they show up for a credentialing exam. Through our collaboration with BioBuilder, we also aim to incorporate this technology into a credentialing tool, providing evaluators with resources they can use to ground their feedback and evaluation with recorded evidence. The hope of this collaboration is that we can dramatically scale the impact of these workforce development initiatives and allow students to provide deeper evidence of mastery of the skills that are required for scientific work in advanced biomanufacturing, enhancing their candidacy.

For instructors, Transfyr’s platform allows them to extend their attention. Trainers still set the standards, review ambiguous or consequential assessments, and coach the learner, but the platform allows them to spend more time with each student’s work, surfaces relevant “moments,” organizes feedback, and tracks performance over time. The intended result is a continuous learning record rather than a one-time pass or fail. Programs can use the same record to improve instruction, compare assessment across cohorts, and show employers what a credential represents.

For future employers, this delivers candidates who arrive ready on day one, backed by verified, objective data proving their physical proficiency rather than just course completion.

Employee Onboarding

While statewide workforce development programs are geared toward teaching core competencies that support a robust biotech workforce, individual companies require more specialized training geared towards specific assays, devices, and workflows that are unique to a given company. As new employees join the company, our goal is that AI-assisted training can accelerate the time-to-impact for new hires. A diagnostics partner of ours shared that new employees spend up to six months “pipetting water” before they’re trusted to handle real human samples. This onboarding time is extremely costly and can present a major bottleneck in scaling operations. Transfyr’s physical AI platform can help employers by capturing a high-fidelity SOP by observing their expert users, and use that to design interactive tools for new employees that provide instant feedback on their proficiency on a given protocol, or tips for troubleshooting, thus dramatically reducing employee onboarding times, and a system of record to capture a given protocol and the associated novice-to-expert continuum operation of the protocol.

Uplifting Humans in an Automated World

While mastering foundational, hands-on bench techniques remains essential, it is only the baseline for the modern laboratory. As massive investments accelerate automated science, routine physical execution will increasingly shift to connected instrumentation and robotics, requiring workforce development programs to evolve in tandem. The scientific workforce of the future must seamlessly bridge the gap between traditional manual skill and the agility to onboard, operate, and troubleshoot automated systems. Without this dual focus, advanced infrastructure risks sitting idle, and training programs risk preparing learners for roles that are already obsolete. Our collaborations with BioBuilder and large organizations focused on employee onboarding integrate physical bench proficiency with advanced laboratory automation from day one.