Try the AI‑powered demos

Free during our launch — no account needed. Pick one and go.

API Factory: find the step that limits output

Our software lays out an API plant from a documented chemistry route, times one batch, and finds the step that limits yearly output. Aspirin example: the reactor sets the pace, for about 598 metric tons a year when the recovered liquid is reused, or about 513 without it (330 days, 85% uptime). All figures are modeled.

What it means for your plant

  • Know what limits output before you spend money.
  • Check whether a change would raise output before you buy it.
  • Once your process data is on file, start from draft batch records and procedures, not a blank page.

What the software does

  • Finds the bottleneck. Names the slowest step and the yearly output that follows.
  • Tests three changes. A second dryer, a combined filter-dryer, or a different uptime.
  • Shows when a change won't help, or hurts. For aspirin, a second dryer adds nothing. For paracetamol, a combined filter-dryer would cut output by about 30%.
  • Shows its sources. Every figure has a source or is marked as an assumption. No AI in the math.
  • Drafts documents when the data exists. Batch record, procedures and bill of materials for aspirin. It refused for paracetamol, where that data was missing.
  • Screens chemicals first. Checks every chemical against a starter list of controlled and dangerous substances, and stops on a high-risk match.

Why build a digital twin

  • See how a plant should run before you build or change it.
  • Give engineering, operations and quality one shared view.

Planning a new line or a debottleneck? Email us the molecule and the step you think limits output.

Code and film by Claude (Anthropic), under our direction. Concept demo, not for GxP use. The twin is a planning model with no live plant data. No human expert has signed it off, and nothing has been validated or run in a plant.

See the 7 steps in the video

Opening: an aspirin structure image is dropped into a molecule box. The engine does not read the image; it works from aspirin's documented, patent-based route. Then the batch:

  1. Charge and heat: acetic acid 622 L, then salicylic acid 1,500 kg.
  2. Acetylation: acetic anhydride 1,552 kg metered over at least 180 minutes; the feed valve shuts itself outside 80–90 °C. This reactor block is the slowest station (20.5 h).
  3. Seeded crystallization: seed 1.96 kg at 78 °C, cool to 12 °C (hard low limit 10 °C).
  4. Centrifuge and wash: acetic acid displacement wash first, then acidified cold water; mother liquor goes to recycle.
  5. Vacuum drying: endpoint by loss on drying plus residual acetic acid by GC.
  6. Milling and sieving, then packaging.
  7. Batch complete (modeled): 1,561 kg packed on the first batch.

Sources: chemistry route based on US patent 2,987,539 (2,731,492 and 3,235,583 cited for comparison); station hours and sizes from the engine's plant model.

Questions? Email info@pharmtech.ai.