Skip to content
TalentXM

Proof of concept · Manufacturing

ChemistXM: Chemical Structures from Scientific PDFs

Extracts molecular structures from PDFs, generates SMILES and enriches each result for chemist review, keeping the source page.

Video and presentation

ChemistXM POC walkthrough4:27
ChemistXM commercial overviewOpen in a new tab

Overview

The problem, and what the POC does.

The problem

Chemical structures are trapped inside PDFs, reports and customer records. Chemists re-key them by hand, different tools produce different identifiers, and records lose the link to their source page.

What the POC does

A PDF dropped in an approved SharePoint folder or attached to a Salesforce record starts a Power Automate flow. A Cloud Run container runs a DECIMER pipeline that finds each structure and generates a candidate SMILES string. RDKit canonicalises and enriches it, and low-confidence results go to a chemist, who accepts, corrects or rejects them.

Outcomes

What it is built to improve.

  • Faster first-pass interpretation of chemical documents
  • More consistent chemical identifiers
  • Every structure traceable to its source document and page
  • Clear review queues, with the chemist making the final call

Technologies used

Request an Agent Pilot

Run Chemical-Document Interpretation on your workflow.

Tell us about your systems and timeline. We will scope a pilot with agreed measures and human decision points.

For example your EHR, ERP, integration platform or Microsoft 365.

More POCs

All POCs
3:23

Manufacturing

Machine Health Monitoring

Learns a baseline for each machine, flags anomalies in telemetry and explains them through a diagnostics assistant.

  • Anomaly detection (Isolation Forest)
  • Microsoft Copilot Studio
5:09

Manufacturing · Finance

AP Vendor Email Assistant

Reads vendor emails, checks the sender's access and finds the answer in finance systems, then replies, drafts for AP review or escalates in Teams.

  • Workato
  • Microsoft Teams
  • Large language model