Invert

Your process model has to earn trust before anyone follows its recommendations. In this clip from the audience Q&A, Jack Prior, Head of Process Monitoring and Data/AI Strategy at Sanofi, explains how predicting a batch before it runs helps build that trust. Masaki Yamada, our Head of Product, adds why you need to know where you are data-light. The September 16 BioPharma Webinars session has both talks: Jack on how to measure whether a network's process data is ready to act on, and Masaki on what it takes to put that data to work. The full recording is in the comments.

Next Wednesday (9/16), BiopharmaWebinars is hosting two talks on the gap between collecting process data and actually being able to use it for monitoring, decision-making, and AI. Jack Prior (Sanofi) on BioMetre, his maturity framework for process monitoring readiness: how 25+ products scored on it, and the pattern across them. Masaki Yamada (Invert) on the AI agents we build for routine bioprocess decisions, like deviations, batch comparisons, tech transfers, and regulatory justifications. Excited to see everyone there! Can't make it live? Register anyway and the recording comes to you. 9:30am ET / 6:30am PT / 2:30pm BST / 3:30pm CEST Register: https://lnkd.in/gsV5vCkB

You rarely catch missing data during the run. It shows up months later, mid-analysis, when a curve looks a little funny and you can't tell whether to trust the result or repeat the run. Our Head of Product, Masaki Yamada pointed Assist at 12 Ambr250 runs. It came back in 2 minutes 36 seconds with three problems: an antifoam value that was just the letter "v", an empty density field, and one offline glucose series that's gone for good. Link to learn more in the comments.

Get the latest jobs and industry news

On many bench-scale reactors, native control software has limited memory, so it downsamples what it shows you... The full-resolution signal exists on the system, but the view you get is partial. Invert reads that raw signal directly over OPC and keeps it at full resolution, together with your offline assays and in the context of the run and its phases. In a CHO run, for example, dissolved oxygen can spike and settle in a short-lived excursion lasting only a few minutes. In a downsampled view, that flattens into a smooth line. When a batch underperforms, an excursion like that could be an important clue, but only if it was captured. If it was averaged away before anyone saw it, root-cause analysis may stretch from days into many expensive months. Invert keeps the full signal live and in context, so you can spot a deviation as it happens and still have the data to reconstruct it later if needed. Read how we do it → https://lnkd.in/gqCRTJb2

What would you ask your data if the answer cost a sentence instead of a week? You can use Assist to fetch information, e.g., “pull up my last 5 filtration runs” or “show me last week’s titers.” But what we’re seeing more and more, is teams asking Assist questions that build on information at the run level: “across our current run campaign, what are the highest performing feed and media conditions, and what should we try next?” A new layer of insight on top of the data you already have. That’s why we built Assist (and why we’re demonstrating it can replicate models in bioprocess research papers…). Read about Invert Assist here - https://lnkd.in/gu9bWW32