4.12.22-Health-Science-Startups-Fathom-Data

ELISABETH MAIDA: Hi, everyone. My name is Elizabeth Maida. And I'm the co-founder and CEO of Fathom. Prior to founding Fathom, I founded a cybersecurity company, which was acquired by McAfee, and worked for many years at Akamai Technologies right here in Kendall Square. I also attended MIT for graduate school and earned dual master's degrees in computer science and engineering systems.

Now, Fathom's mission is to simplify access to biomanufacturing data. Our vision is to aggregate and analyze data from throughout the manufacturing lifecycle to enable real-time visibility.

In biopharma, a lot of attention has been paid to how AI and ML can enhance drug discovery. But not as much focus has been put on the application of those technologies to manufacturing data. In our experience working with pharmaceutical companies, we saw how much of this critical data was still trapped on paper and in legacy ERP systems.

Now, as I think everyone in this room is aware, there's been a fundamental shift in the biopharmaceutical industry from chemical compounds to biologics and cell and gene therapy. And it has profound implications for how you're able to actually characterize the output of these processes and assess their quality.

With traditional chemical compounds, there are standard tests available that can measure how much of each element is present and detect any impurities. But with biologics and cell and gene therapies, you really have to have insight and visibility into the process by which these therapies are manufactured. Right now, the majority of this data is kept in disparate systems and very siloed from each other, despite how critical this data and information is in determining the efficacy of the therapies.

Now, while Fathom's long-term vision involves automatically integrating directly with the manufacturing equipment, our initial focus is on extracting this data from documents. Fathom's OCR and NLP accurately extracts this data and then provides automated validation, checking to confirm that handwritten calculations are correct, identifying the absence and presence of initials and dates, and making sure that the data conforms to scientific realities.

Now, these may sound like old problems. But existing tools do not work out of the box. This technology really requires a domain-specific approach, as scientific language is not natural language. Existing solutions have been built for invoice statements and financial data, not for tracking cell viability or determining whether a signature is actually accurate and would be accepted for compliance purposes.

With Fathom's approach, we have layered subject matter expertise in with our unique, proprietary OCR and NLP technology to ensure the accuracy in the trending of the extracted data. As an example, we've been working with an allogeneic CAR-T therapy company who gets donor cells and genetically modifies them so that it can be injected into other patients without triggering an immune response. It's a 19-day manufacturing process. And they often send samples of the cells out to other labs, get results back, and handwrite them into the batch record. Their biggest issue is human error. And if they can catch those issues earlier, they can limit the number of cells that may be at risk of dying.

Now, this can be up to $2 million per batch. But far more critical than the actual financial implications, these are life-saving therapies. In working with Fathom, this company was able to reduce the time required by the quality teams for reviewing the batch records by 80%, from 35 days down to about seven days. In addition, we generated critical data that manufacturing sciences could use to trend critical parameters, analysis that is required by regulatory authorities for compliance.

Now, our focus is not limited to batch record data. We are also working with cell and gene therapy companies to automatically collect and trend environmental monitoring data, correlating non-viable and viable data, overlaying microorganism information to identify location and seasonal-based trends, and trending multiple years of viable data for compliance purposes. This has allowed our customers to respond to due diligence requests from potential collaborators and accelerate the batch release process, reducing the amount of time required to gather that data to less than 30 minutes. This is also a great example of how Fathom is incorporating data from multiple different sources-- in this case, documents, lab results from external parties, and actual system instruments, the particle monitoring systems.

We're working with these companies as part of our design partnership programs, where they get early access to our product and can provide feedback that's incorporated directly into our roadmap. We're actively looking for other customers and companies in the biopharmaceutical industry, especially those that are interested in automating their manufacturing data. If this is of interest, please feel free to reach out to liz@fathom.one or come by our booth during the startup showcase immediately following this. Thank you.

[APPLAUSE]