Data Science · Nürnberg

Priyadarshini M D

Master's student in Data Science at FAU Erlangen-Nürnberg. At Siemens I built the pipelines and Power BI dashboards behind cloud cost reporting. In my own projects I use real German electricity prices to answer practical questions about cost, forecasting and planning.

Ex-Siemens working studentSQL · Python · Power BI5 open-source projects
Germany 2025 · power price by hour · ct/kWh

Projects

Each one starts with a question. The code, data sources and full write-up are on GitHub.

Hospital medicine ordering

Is the most accurate forecast also the best one to order with?

−57%ordering cost, with availability up from 84% to 97%

I forecast daily demand for six medicines and checked which forecast led to the cheapest orders without shortages. It was not the most accurate one. Planning with the uncertainty of the forecast made the difference.

Forecast error compared with the annual cost of ordering decisions for five methods
scikit-learnQuantile regressionStochastic LP
Medicine supply network

How do you keep 15 hospitals supplied when a warehouse can fail?

6%of weeks with shortages, down from 47%

I modelled where to open warehouses and how much safety stock to keep, and tested every plan on 200 scenarios with uncertain demand and random disruptions.

Expected cost and service level of deterministic, stochastic, fair and robust supply plans
Stochastic MILPRobust optimizationHiGHS
Home energy optimizer

When should a home battery and an electric car charge?

−11.5%yearly energy cost vs. the standard battery rule

Every day at 13:00, when tomorrow's prices are published, my planner decides when to charge and when to use stored power. It runs on real 2025 prices and Nürnberg weather.

One week in May 2025 comparing the standard battery rule with smart control, with real prices, PV output and battery level
PythonMILPForecasting
Heating network dispatch

How should a small heating network run its machines, hour by hour?

−24%operating cost, and 25% less CO2

I scheduled a heat pump, gas engine, boiler and heat storage for one winter week, with start-up costs and minimum run times, and compared it with a simple hour-by-hour controller.

Optimal heat dispatch over one winter week with heat pump, CHP, boiler and storage
Unit commitmentMILPEnergy systems

How I work

Start from the decision

A forecast or a dashboard is useful when it helps someone decide. In my medicine project, the most accurate forecast did not lead to the best orders.

Check the data first

My pipelines test for missing hours, impossible values and totals that don't reconcile. At Siemens I traced gaps in carbon reporting back to their source.

Make it easy to use

I build for the people who read the numbers: Power BI reports for finance teams, a live dashboard, and READMEs anyone can follow.

Experience and education

04/2026 – 09/2026

Working Student, FinOps and Data Engineering

Siemens AG, Erlangen
  • Built pipelines that load AWS cloud cost data into internal systems through APIs
  • Created Power BI dashboards and KPI reports for finance, cloud and data engineering teams
  • Traced gaps in carbon emissions reporting back to account mappings and source data
Enrolled

M.Sc. Data Science (AI and Machine Learning)

Friedrich-Alexander-Universität Erlangen-Nürnberg
01/2024 – 02/2025

Software and Data Intern

The Megha Foundation and Bharatiya Yuva Shakti Trust, Chennai
11/2022 – 12/2023

Backend Developer

Market Simplified India, Chennai
  • Backend features for the Arihant Capital trading app with Java, Spring Boot and SQL
2019 – 2023

B.E. Computer Science and Engineering

RMD Engineering College, Anna University, Chennai · CGPA 8.9/10

Skills

Data

SQL, Python (pandas, NumPy), ETL pipelines, data quality tests, Snowflake, Git, AWS

Analytics and BI

Power BI (DAX, Power Query), Looker Studio, KPI reporting, budget vs. actual

Modelling

Machine learning with scikit-learn, forecasting, linear and mixed-integer optimization

Languages

English: fluent
German: B1
French: basic

Let's talk

I'm based in Nürnberg. The easiest way to reach me is by email.

priyadarshinimd253@gmail.com