AI Email Interpreter with NLP

Status: Closed
Prize: $400
Entries Received: 12
Winner: ZahraAcc

Winner

Contest Brief

You are building an AI text interpreter for an IT parts company.

You will build an email interpreter that utilizes Natural Language Processing (NLP). This system should be capable of identifying the correct part number requested in emails (based on the part type asked for, e.g. battery, LCD, hard drive, etc), and get the bill of materials to find this part number based on serial and/or model numbers provided by the senders.

Key Tasks:
- Develop an AI model that can comprehend and interpret email content.
- Implement a lookup system against manufacturer bill of materials data sources.
- Ensure the AI can reference various data sources including External APIs, CSV/Excel files, HP / HPE Partsurfer, and Lenovo Parts Lookup.

Integration:
- The AI system needs to be integrated directly with Office 365 emails and read all inbound emails. It will respond to the sender directly if it has the part number, or move the email to a folder if not.

===
ACTION PLAN AND WHAT IS REQUIRED
Phase 1: (this contest) - extract that data and populate a spreadsheet with the sender email and name, product requested, quantity requested, product model and product serial. Send this data to HPE Partsurfer and find the corresponding part from their Bill of Materials information.
Example search at HPE:
https://partsurfer.hpe.com/Search.aspx?SearchText=CZJ436022W
--> Product Number = Model number
--> Advanced (Unique) BOM tables show all the items specific to this exact serial number (most accurate source)
--> General (Model) BOM tables are a fall back if the exact item is not present in the Unique BOMs (e.g. maybe Unique only has a 4GB RAM module and we want the part number for a 16GB RAM module - go to General to try find that)

NEXT STEPS - i.e. not required now, but scope so you can do these
Phase 2: (winner will get this piece of work) - send that data to our 4 next biggest manufacturer sites and establish what the part number is based on their bill of materials information. Some are based on APIs, and some on ugly Excel files.
We will then push this data to our internal API to get the price for this customer so you can construct and accurate reply to the sender.

Phase 3: apply similar logic to interpret images
====

WIN CRITERIA:
The best entry will be:
1) Most accurate in identifying the required data points with various email structures
2) Most accurate in finding the correct part from manufacturer data source based on those data points
3) Able to explain development plan for phase 2 + 3