Skip to main content

Start with an idea

Novasign AI Assistant is your optional copilot in Novasign Studio. It works as a sparring partner, a teacher and a worker. You can improve your work with it step by step, ask what was done and why, and give it instructions so you do not have to do everything yourself. It can help with a single step of your project or take you all the way from an idea to a digital twin.

Book a demo

From a file you drop to a project you can run

The AI Assistant is embedded in Novasign Studio. Drop your files into the chat, say what you want to find out, and it works through the project with you.

Novasign Studio / AI Assistant
  1. You
    • online_bioreactor.csv
    • vcd_offline.csv
    • product_titer_offline.csv

    Create a dataset that is ready to model VCD and product titer based on temperature and pH.

    14:02
  2. AI Assistant

    The three files become one dataset in a new project, Fed-Batch Study. Before I join them, please check the units.

    ColumnHeader UnitStored As
    tempnone°C
    feed_volumemLmL
    vcd10⁶ cells/mL10⁶ cells/mL
    titerg/Lg/L

    The column temp has no unit. From the size of the values, 32.8 to 37.2, I assume degrees Celsius.

    • Use suggestion (°C)
    • Define unit
    14:03
  3. You

    Use suggestion (°C)

    14:08
  4. AI Assistant

    Done, temp is stored in °C. I see a feed column (feed_volume in mL) and glucose measured at several time points. Is the feed a glucose feed? If so, what was the glucose concentration in the feed?

    14:09
  5. You

    Yes, glucose was added as a constant feed with a concentration of 400 g/L.

    14:14
  6. AI Assistant

    Thank you. Here is a short summary.

    • You ran 12 fed-batch runs and varied temperature at 33, 35 and 37 °C and pH at 6.8, 7.0 and 7.2.
    • Glucose was measured 4 to 5 times per run, and VCD and titer once a day.
    • I joined the offline measurements to the online data by run and time, set temp to °C and added the glucose fed from feed_volume and the 400 g/L feed.
    • The dataset is ready to model VCD and titer from temperature and pH.
    Open the dataset in Studio
    14:16
Message AI Assistant

The AI Assistant can make mistakes. Please verify responses.

  1. It asks what you are trying to do

    Your goal decides what happens next, so the AI Assistant asks for it instead of guessing. Pick one of the options it offers or say it in your own words.

    AI Assistant / Your Goal

    What do you want to do with this data?

    • Explore Data

      See what was varied, how often each variable was measured and where values are missing.

    • Curate Data

      Join online and offline data, align them on one time base, convert units and fill the gaps that should be filled.

    • Model

      Identify a suitable model and get suggestions on how to configure it.

    • Simulate

      Find out which conditions that have never been run are worth investigating for your objective.

  2. It investigates your data like a data scientist

    Ask about your data and it queries every row to find gaps, constant columns and values that stand apart. It shows what it finds in charts and tables and names the preparation steps your data needs, without changing the data itself.

    Fed-Batch Study / 3 Files

    • time

      Type
      numeric
      Unit
      h
      Observed Range
      0–336
      Measured
      online

    • temp

      Type
      numeric
      Unit
      none
      Observed Range
      32.8–37.2
      Measured
      online

    • pH

      Type
      numeric
      Observed Range
      6.78–7.23
      Measured
      online

    • feed_volume

      Type
      numeric
      Unit
      mL
      Observed Range
      0–420
      Measured
      online

    • glucose

      Type
      numeric
      Unit
      g/L
      Observed Range
      0.4–6.8
      Measured
      4 to 5 per run

    • vcd

      Type
      numeric
      Unit
      10⁶ cells/mL
      Observed Range
      0.3–24.6
      Measured
      daily

    • titer

      Type
      numeric
      Unit
      g/L
      Observed Range
      0–3.4
      Measured
      daily

    • run_id

      Type
      categorical
      Observed Range
      12 levels

    What It Sees

    • Fed-batch cultivation
    • Temperature and pH varied at three levels
    • Constant glucose feed
    • VCD and titer measured offline
    • Temperature has no unit

  3. You agree on a plan before anything is built

    From designing experiments to importing and curating data, building a model and setting up a digital twin, the AI Assistant tells you what it would do. It knows which nodes, functions, prebuilt models and digital twins Studio has and picks the ones that fit. It creates the project and its folders as you agree. You can keep refining the plan in the chat or accept it and inspect the result in Studio.

    AI Assistant / Proposed Plan
    1. Import and CurateRead the three files, join VCD and titer to the online data, convert temperature to °C and add the glucose fed.5 nodes
    2. ModelTrain a reactor model for VCD and titer from temperature and pH, and select the best fit.2 nodes
    3. Digital TwinSimulate combinations of temperature and pH that were never run and show the results.2 nodes

  4. It sets everything up in Studio

    The AI Assistant builds what you agreed on in Studio itself. It chooses each node, fills in its settings and connects it. Everything it adds is marked as a suggestion, and a summary lists the changes so you can inspect them first or accept them all at once.

    Novasign Studio / Fed-Batch Study / Model and Process Simulator

Or let it help with a single step

You do not have to hand over a whole project.

  • ExplainAsk what a step does and why it is set up the way it is.
  • SuggestAsk what it would change, and why, before anything is touched.
  • ActLet it make the change. It arrives as a suggestion that waits for you.

What the AI Assistant draws on

Novasign Studio is built from the ground up to be driven by agents. The AI Assistant works with every part of it through the same services the Studio interface uses, and brings together what Studio offers, what the language model understands and what Novasign has learned from years of bioprocess modelling.

Inspect, adapt and accept

Every change the AI Assistant suggests shows up in Studio marked in yellow and waits for you. Open it and compare it with the original, adjust it yourself or ask for another version, then accept or reject it. Colleagues in the same project do not see your pending suggestions. Working with it is a conversation toward your goal, and nothing enters a workflow until you accept it.

Novasign Studio / Fed-Batch StudyMBJK+1ConnectedAI Assistant
3 Suggestions PendingVisible to You Only

Working with suggestions is meant to feel natural. Each one opens next to the original so you can compare the two, you can adapt it yourself or ask the AI Assistant for another version, and you accept or reject it one at a time or all at once. Nothing overwrites a colleague’s work.

How Novasign Studio works with AI

The AI Assistant is optional in Novasign Studio. If you use it, it works with a hosted model such as GPT or Claude or with an open-source or open-weight model inside your own deployment, and you choose which one.

Your Deployment
Novasign StudioYour Projects and Data
Stored in Your Deployment
  • Projects and Workflows
  • Datasets
  • Runs and Results
AI AssistantOptional
Either One
Model in Your DeploymentOpen Source or Open Weight

Prompts and their context stay inside your deployment.

Either One
Hosted ModelSuch as GPT or Claude

Prompts, with any project content or data samples they carry, go to the provider you choose, if your policy allows it.

Where the model runs. With a model inside your deployment, no prompt, no project content and no data sample leaves it, and the knowledge index stays local too. With a hosted model, prompts and the project content or data samples they carry go to that provider, so use one only where your policy allows it.

Colleagues’ work. A suggestion cannot be accepted while a colleague has that node open, or after the node has changed since the suggestion was made, so nothing overwrites their work.

Organizing a project. It happens as you ask. Creating the project and its folders and naming and moving items is filing rather than science, so it is not put to you as a suggestion. Deleting is not something the AI Assistant can do.

Runs and publishing. Starting a run always asks you first. Accepting a suggestion does not publish anything either, because publishing stays a separate step that only you take.

Bring your data and your AI policy

Novasign is happy to do an IT assessment with you, so that it is clear how the AI Assistant fits your infrastructure and your policies.

Book a demo
Related