SAP consultant to FDE: master data and integrations for AI work

Dear Ramesh, about your eight years in MM and SD

You told me you have done three S/4HANA rollouts, spend your weeks writing functional specifications, and are the person people call when IDocs pile up in error status. You also said you are tired of being "the config person" and want to build things. Here is how I would think about it.

Start by listing what you actually handle, because it is more relevant than you think:

• Functional specifications that turn a business requirement into fields, rules and exceptions.
• Master data: material master, vendor master, customer master, pricing condition records, and the endless cleanup before go-live.
• IDoc monitoring and reprocessing (WE02/WE19, BD87), which is really integration debugging.
• Test scripts and UAT sign-offs, with defect logs.
• Cutover plans with dependencies, owners and a rollback point.

A client that wants an AI assistant to answer "where is my order" or draft purchase requisitions will need someone who knows that half the vendor master is duplicated and that the order status lives in three tables. That person is usually missing from AI teams.

Your first project: find duplicate vendors before an AI trusts them

Take the problem you know best. Before any assistant is allowed to create a purchase document, it has to pick the right vendor. The query below finds likely duplicate vendors in a synthetic vendor table, by normalising names and matching on tax ID or on normalised name plus city. It runs in SQLite as written.

-- Synthetic vendor master extract (hypothetical data)
CREATE TABLE vendor (lifnr TEXT, name TEXT, city TEXT, tax_id TEXT, created TEXT);
INSERT INTO vendor VALUES
 ('100045', 'Shree Ganesh Packaging Pvt. Ltd.', 'Noida', 'TX-7781', '2019-04-02'),
 ('100912', 'SHREE GANESH PACKAGING PVT LTD', 'Noida', 'TX-7781', '2022-11-15'),
 ('101230', 'Shri Ganesh Packaging', 'Noida', NULL, '2024-01-09'),
 ('100377', 'Apex Fasteners', 'Pune', 'TX-2210', '2020-06-30'),
 ('101501', 'Apex Fasteners.', 'Pune', NULL, '2025-02-12'),
 ('100600', 'Apex Fasteners', 'Chennai', 'TX-9054', '2021-08-01');

WITH norm AS (
  SELECT lifnr, city, tax_id, created,
         TRIM(REPLACE(REPLACE(REPLACE(REPLACE(UPPER(name), '.', ''),
              'PVT LTD', ''), 'PRIVATE LIMITED', ''), '  ', ' ')) AS key_name
  FROM vendor
)
SELECT a.lifnr AS keep_vendor, b.lifnr AS suspect_duplicate,
       CASE WHEN a.tax_id = b.tax_id THEN 'same tax id'
            ELSE 'same name and city' END AS reason
FROM norm a
JOIN norm b
  ON a.lifnr < b.lifnr
 AND ( a.tax_id = b.tax_id
       OR (a.key_name = b.key_name AND a.city = b.city) )
ORDER BY a.lifnr, b.lifnr;

It returns two pairs: 100045/100912 on the same tax ID, and 100377/101501 on name and city. Notice what it does not catch: "Shri Ganesh Packaging" differs by one letter, so it slips through, and the Chennai Apex is correctly left alone as a different branch. Write that up honestly. Then add a second step in Python with a fuzzy-match score, and state the threshold you chose and why. An interviewer will learn more from your explanation of the Shri/Shree miss than from a perfect-looking result.

The point of the project is the sentence you can say at the end: "An assistant that creates purchase requisitions must check this first, or it will book spend against a duplicate vendor."

What you will find hard, honestly

Three things, Ramesh. First, writing and testing code every day; configuration is not the same muscle, and the first months will feel slow. Second, model behaviour: you will need to measure how often an assistant picks the wrong vendor or invents a status, and report it as numbers rather than impressions. Third, infrastructure: containers, cloud deployments and access control for model endpoints. The FDE for Professionals track at /courses/fde-professionals is built around these gaps for experienced people; its schedule is still proposed, so check the current status with a counsellor.

Integration or solutions engineer as the landing role

I would not aim at "AI engineer" on day one. Look first at roles where SAP knowledge is the entry ticket and engineering is the growth: integration engineer or solutions engineer on an AI product that connects to ERP systems, technical consultant on an AI-for-procurement or AI-for-finance product, or a forward deployed role at a company whose clients run SAP. In each, you are hired partly for what you already know, which gives you room to grow as an engineer.

What you might say when they ask "why leave SAP?"

"I am not leaving SAP so much as moving one layer up. In three rollouts I saw that the hard part was never the configuration; it was master data and integrations failing quietly. I started writing SQL and Python to find those problems before go-live, including a duplicate-vendor check I can walk you through, with the cases it misses. AI assistants that act on ERP data will fail the same way unless someone who understands the data is in the room. I want to be that engineer, and I have spent the last months building the coding and evaluation skills to do it properly."

When you are ready, compare your resume against the role with the skill check. And read the FDE hub so you can describe the job in your own words.

Good luck with it.