TetraScience

At our live Technical Showcase last week, Scientific Solutions Partner James Davies, PhD picked an instrument TetraScience had never worked with before and built a full data flow for it in one sitting. The instrument was a Molecular Devices FLIPR fluorescence imaging plate reader. The build covered the whole loop: plate layout sent from a Revvity Signals ELN to the instrument, raw results routed back through Tetra OS, a new IDS schema representing the data in a vendor-neutral format, calculated results written back to the ELN, and a visualization app for reading the dose-response curves. He used Claude Code with Tetra plugins installed, and none of it was transcribed by hand. Any AI assistant can generate code from a blank prompt. What it can't bring on its own are the conventions Tetra OS already encodes: how a plate reader schema needs to be structured, and which components already cover most of what a lab integration requires. Those conventions came from years of prior deployments. They're what let this build move fast without leaving something the next engineer has to redo. James walks through the build, and why grounding AI tooling in well-modeled data matters for this kind of work, on the blog. Short excerpt below on why using Claude on Tetra OS beats reasoning over non-AI-native data. https://lnkd.in/ewdesyYx

Your laboratory instruments are generating data all day. Getting that data into a usable format is hard enough. But then you have to keep those pipelines running reliably across dozens of instruments and multiple sites. That's where everything usually falls apart, because when a DIY connector fails quietly, you don't find out for weeks. Not with the Tetra OS platform. Our latest updates extend useful new capabilities into the operational layer where the real work happens (and we added a new Dark Mode to save admins' tired eyes. -- Real-time alerts catch silent connectors, throughput drops, latency spikes—within minutes. -- A global search finds any ingestion path across your fleet without digging through spreadsheets. -- Live job monitoring so you can see what's happening without bugging your account team. -- Idle apps shut themselves down so you're not paying for compute that's just sitting there. Noticing that data is missing at review time is no one's idea of a best practice. These alerts surface it in minutes. Full release details: https://lnkd.in/ghbwKVtb

Join TetraScience's Sept. 17th showcase to learn how Tetra OS is applying AI agents to accelerate lab automation workflows connecting instrumentation, LIMS, and ELN systems.

Building a new scientific data integration used to mean weeks of back-and-forth between scientists and data engineers: requirements sessions, schema design, pipeline work, ELN wiring, testing. One instrument or one use case could take up weeks of calendar. We've been cutting that down dramatically using AI coding assistants alongside Tetra OS. In two weeks, co-founder Siping Wang and solutions architect James Davies, PhD are going to show exactly what that looks like, start to finish. Starting from a real lab dataset, they'll gather requirements, build out the data schema, push data to an ELN, and test the result end to end. All from a local IDE, using Tetra MCP and Tetra skills on top of Tetra OS. Two sessions: September 17 at 9:00 AM ET (3:00 PM CEST) and 12:00 PM ET. Register: https://lnkd.in/eMAa2xsz

When Boston-based TetraScience was planning its European headquarters, the team needed a location close to the pharma companies and scientific teams its platform was built to serve, with access to the specialized talent needed to grow. Read more about the company’s international expansion journey and why Basel Area was the right fit.

We're thrilled to welcome Marco A. Zampini, PhD to TetraScience as a Scientific Data Architect. Marco brings deep expertise in quantitative imaging and signal processing, with eight years of experience designing, validating, and deploying advanced MRI techniques in preclinical and clinical settings. His background spans pulse sequence development, reconstruction algorithms, and automated data processing pipelines—work that required him to move seamlessly between instrument design, algorithm development, and real-world implementation across animal models, phantoms, and human studies. That combination of signal engineering and practical data workflow experience is exactly what matters here. Marco understands the full lifecycle of scientific data from instrument through analysis. He's built systems that turn raw, complex signals into structured, usable outputs. That's the work TetraScience does across biopharma. Here's what Marco had to say about joining us: "I'm joining TetraScience after eight years immersed in imaging and signal analysis in the preclinical world, the kind of work where you learn to trust data only after you've earned that trust the hard way. During my PhD and since, I've been drawn to pushing accelerated MRI techniques from a promising idea to something a scientist could rely on, and that process taught me just how much careful engineering, iteration, and validation sits behind every trustworthy signal. What excites me about this role is the chance to bring that same instinct to the much bigger canvas of biopharma R&D, for turning complex data into something people can act on. I can't wait to see how these principles hold up (and where they'll need to evolve) at that scale, and to help push the mission forward." Want to join Marco in building the operating system for scientific intelligence? Open roles here: https://lnkd.in/eD_bZu8Z

I'm excited to be joined by Jenn Medellin for our next Tetra Technical Showcase on August 13! We're doing something a little different this time: building a real scientific app, live, from an empty folder to a published app tile, in about 30 minutes. Jenn Medellin will build a 96-well QC plate review screen on Tetra OS using plain-English prompts to an AI coding assistant. What makes this work isn't just that AI can write code fast. It's what it's writing on top of. The data underneath is already governed: schemas, lineage, and structure in place before a single line of app code gets written. That's the difference between an app that looks like it works and one you can actually trust with real lab data. If you've ever needed a custom view on your data and didn't want to file a ticket and wait, this one's for you. August 13 | 9:00 AM ET / 12:00 PM ET Register: https://lnkd.in/gs8VQckc

Scientists are building analytics apps with AI. Informaticists without front-end experience are, too. The gap between "I have an idea" and "I have a running app" is closing fast. But a fast-built app sitting on top of uncontrolled data is a liability. It looks like progress until it doesn't hold up. Building with AI only works when data schemas are defined, lineage is tracked, and governance is built into the platform before anyone writes a single line of application code. That way, the AI assistant isn't guessing which dataset to use or inventing numbers to fill gaps. It's building on data the organization already trusts. That's what we designed Tetra OS for. Every app built on the Tetra Data Foundry reinforces the schemas, taxonomy, and ontology underneath it. Each subsequent build is faster and more trustworthy than the last. We're walking through what this looks like in practice at our August 13 Technical Showcase — scaffolding, building, and publishing a working scientific data app on Tetra OS, live, in 30 minutes. Read the full post: https://lnkd.in/eXtRxgau Register for the showcase: https://lnkd.in/erVZNnxU

From raw biologics data to trusted, decision-ready insight—faster, more consistently, at scale. A global biopharma company automated one of its biologics assay workflows with TetraScience. Three months in: - 12,000+ AI-ready data points generated (structured and queryable) - Analysis speed <10 minutes per sample instead of an hour - Cycle time: 78% reduction across assay types - Manual effort: 65% removed from each data review cycle When data stops requiring manual extraction, cleaning, and transcription, scientists stop doing janitorial work and start doing... science. Structured data also compounds: cross-program learning and AI model retraining become possible at scale. This is what closing the "data utility" gap actually looks like.

We don't know a ton about yoga pants, but this op-ed from lululemon's former CIO resonated with us in a big way. Like a lot of senior executives, she's seen too many enterprise AI demos that look magical and then stall out before any real deployments. Why? Companies skip the foundational work of harmonizing the data and designing systems to make consistent, repeatable decisions at scale. It's slow work. It's expensive. It requires committed people. The only way to build real, compounding intelligence with AI means solving the foundational problems first. That's what we do -- except in science. We leave the fashion business to the experts. Read the full piece → https://lnkd.in/ekJRPXW3