Skip to content

Saba Salehi

Data Analyst · Process Mining & Automation

Clean Data.
Clear Processes.
Measurable Impact.

Celonis · PQL · SAP · Process automation · Applied GenAI

Greater Munich Metropolitan Area, Germany
01.

About Me

I'm a data analyst working in process mining. Companies usually know how their processes are meant to run. The data tends to disagree.

Finding where, and what it costs, is the job. Day to day that's Celonis and PQL against SAP data.

I'm finishing a bachelor's at Technische Hochschule Augsburg and work in English and German.

Why

Teams rarely lack data about their processes. They lack an honest view of what the data already says. That gap is where I work.

How

Start from the event log, not the flowchart. Quantify the deviation, agree the metric with the process owner, then automate the reporting around it.

Connect

Open to projects and collaborations: analytics, data engineering, process automation, and applied GenAI, in manufacturing, logistics, or finance.

Technologies

  • Celonis
  • PQL
  • SQL
  • SAP HANA
  • Python
  • PySpark
  • Power BI
  • Tableau
  • Excel
  • Power Automate
  • YAML

Languages

  • PersianNative
  • GermanWorking proficiency
  • EnglishFluent

Process domains

  • Order-to-Cash
  • Purchase-to-Pay
  • Production
  • Inventory
  • On-time delivery
  • Supply chain

Certifications & courses

Claude Certification

Anthropic

In progress

Celonis Foundations

Celonis

Write PQL Queries

Celonis

Build Analyses

Celonis

Review and Interpret Analyses

Celonis

Build Knowledge Models

Celonis

Data Integration

Celonis

Build Action Flows

Celonis

Object-Centric Process Mining

Celonis

Currently learning

  • Working towards the Anthropic Claude certification, and reading closely on where LLMs are and are not dependable
  • Running open models locally to understand quantisation, context limits, and what a model costs to serve
  • LLM evaluation: ground truth and field-by-field scoring, instead of judging model output by eye
  • Finance process mining: accounts payable and receivable, so far from the Academy side rather than shipped work
Saba Salehi
02.

Experience

Data Analyst @ manroland Goss web systems GmbH

Dec 2024 – Present · Working student · Augsburg, Germany

  • Build the Celonis analyses that show where production, order and procurement processes leave their intended path, writing PQL against SAP data and conformance-checking the result against how the process is supposed to run.
  • Maintain the Studio views people actually open, and the definitions behind them: business entities, KPIs and the relationships between SAP tables, configured in the visual editor and in YAML where the editor stops.
  • Translate SAP business logic into PQL, reading the ABAP calculation rules behind the revenue and margin figures.
  • Manage the data pools, jobs and transformations behind those analyses, so a result can be reproduced months later instead of re-derived from memory.
  • Cover the order-to-cash and procurement processes end to end: revenue and margin development, order intake, on-time delivery, procurement savings, and stock and consumption across both manufacturing sites.
  • Automate recurring reporting with Action Flows, error handling included, so process owners get the same numbers on a schedule instead of on request and the cases that need a human get flagged.
  • Work with process owners across production, purchasing and logistics to turn what they need into a metric definition that survives the conversation, not just the query.
03.

Projects

Where the order process actually goes wrong

Planned

Process mining on 1.5M events of real purchase-order handling: discovering the model, conformance checking against the intended path, and putting a euro figure on the rework loops.

CelonisPQLSQL

Testing a GenAI assistant that reads invoices

Planned

An evaluation harness for LLM field extraction from messy finance documents: ground truth, accuracy scoring, and a catalogue of what makes the model wrong.

PythonLLM evaluationPrompt engineeringPower Automate

Machine data to dashboard

Planned

An end-to-end pipeline from raw sensor telemetry to a monitored dashboard: ingest, transform, schedule, alert.

PythonPySparkSQL

Reinforcement learning for job-shop scheduling

Planned

A scheduling agent measured honestly against the heuristics it claims to beat, written up with the experimental record that makes the result reproducible.

PythonPyTorchReinforcement learning
04.

What I Do

Process Mining

Rebuilding how a process actually ran from its event log, then showing where and how often it leaves the intended path.

  • Event log extraction & modelling
  • PQL analyses in Celonis
  • Conformance & variant analysis
  • Bottleneck identification

Data Analysis

Getting the underlying data into a state where the answer is trustworthy, using SQL and Python against source systems.

  • SQL against SAP HANA
  • Python & PySpark analysis
  • Data quality checks
  • Ad-hoc investigation

Dashboards & Reporting

Turning a finished analysis into something a process owner can open on a Monday morning and act on.

  • Power BI & Tableau
  • Celonis Studio views
  • Knowledge Models & KPIs
  • Stakeholder write-ups

AI & Automation

Closing the loop, so an insight triggers work in the source system instead of sitting in a report nobody reads twice, and using language models where they genuinely beat a rule.

  • Action Flows
  • Power Automate flows
  • LLM extraction & evaluation
  • Alerting on deviation

05. Got something in mind?

Let's talk.

Whether it's a role, a question about one of the projects, or a process you suspect is costing more than it should. I'd be glad to hear from you. Send a message here, or reach out on LinkedIn. I usually reply within a couple of days.