Meet Let’s Connect

Data in Detail

Data scientist and working student at Siemens Healthineers. I build machine learning models, NLP pipelines and AI agents that turn messy data into decisions.

  • Open to data and AI roles
  • Based in Erlangen, open to Nuremberg, Munich, Berlin, Frankfurt

About Me

Hi There!

I’m Meet Goel, a data scientist who turns raw numbers into clear decisions. Since 2021 I’ve grown from building web apps to training models, working daily with Python, {SQL} and the craft of turning data into stories. Today I build AI agents and map enterprise systems at Siemens Healthineers.

Want to know more about me?

Machine Learning: View ML projects NLP and LLMs: View NLP work Data Analysis: View analyses AI Automation: View Siemens work All Things Data: View all projects

Experience

Learn, Build, and Deliver.

Industry work at Siemens Healthineers, internships and leadership roles, most recent first.

Siemens Healthineers

Erlangen, Germany, on-site

Feb 2026 to present

Working Student, Service Management and Operations

May 2026 to present

I turn operational data into AI-powered tools that Service Management Operations uses to monitor, report and decide.

  • Ticket-to-Talk, an AI agent for incident insight. Designed and built a Microsoft Copilot Studio agent that reads raw incident exports from the SHARP ticketing system and returns a filterable KPI dashboard with an incident summary, so the team sees service performance at a glance instead of analysing exports by hand.
  • Automated management reporting. Extended the agent to generate presentation-ready PowerPoint decks in the team’s own template, replacing manual slide preparation with a repeatable, one-step report.
  • Interface Catalogue, one source of truth. Built a Power Apps application on a 13-table relational data model in Azure SQL Server, with bulk Excel ingestion through Power Automate, giving the department a single, trusted record of technical interface metadata across enterprise systems.
  • Governed data layer. Encoded business rules and data-quality controls in T-SQL stored procedures (existence checks, auto-creation logic, global search), so every app built on the data inherits the same validated logic.
  • AI enablement. Automated communication documentation with a Copilot Studio agent and Power Automate as part of the department’s AI enablement.
Copilot StudioAzure SQL ServerT-SQLPower AppsPower AutomatePowerPoint

Intern, IT Order-to-Cash

Feb 2026 to May 2026
  • Designed and deployed multiple Generative AI agents in Microsoft Copilot Studio, translating department process requirements into working agent solutions, including agents that turn complex documents into structured insights.
  • Modelled the IT Order-to-Cash interface architecture in SAP Signavio, mapping how business processes flow across more than 50 integrated systems around the ERP.
  • Evaluated Power BI, LeanIX, SAP Signavio and Power Apps with business and data experts before choosing SAP Signavio as the best fit.
  • Completed SAP AI workshops on Joule agents, SAP BTP AI Core, AI Launchpad and SAP Build, and worked within the Azure DevOps lifecycle.
Copilot StudioGenerative AISAP SignavioAzure DevOpsSAP BTP
Dec 2023

Winter Training, Full Stack Development

Indian Institute of Technology, Kanpur

Web development, databases and server-side scripting through hands-on projects, including an expense tracking system.

Oct to Nov 2023

Intern, Full Stack Development

Open Weaver, remote

Built and deployed an e-commerce site in Open Weaver Studio with registration, catalogue and order workflows, SMTP emails and Razorpay payments.

Aug to Oct 2023

Intern, Python with Data Structures and Algorithms

YBI Foundation, remote

Python programming, data structures, algorithms and first machine learning notebooks.

Oct to Nov 2023

Intern, C++ Programming

TechnoHacks EduTech, remote

C and C++ programming tasks and problem solving.

Sep to Nov 2023

Intern, Android Development

EduSkills Foundation, remote

Android app development in Kotlin.

Jun 2025

Event Coordinator Volunteer

GUS Germany, Berlin

Venue setup, promotion and attendee experience for the GGG Summer Career Fair. Earlier: Core Committee Coordinator for Hack-A-Thon 2K24 and Event Coordinator for INVENT 2K23 at IMS Ghaziabad.

Education

Study, Research, and Grow.

From commerce in school to computer applications and a master’s in data science and AI.

Sep 2024 to Sep 2026

MSc Data Science, AI and Digital Business

Gisma University of Applied Sciences, Potsdam

  • Thesis on cross-lingual and few-shot offensive language detection with XLM-R and Qwen2.5.
  • Built Cinemate, an NLP movie recommender deployed on AWS EC2, for the Big Data and AI business project.
  • Projects in regression, classification, time-series forecasting and exploratory analysis in Tableau.

Jul 2021 to Jul 2024

Bachelor of Computer Applications, grade A+

IMS Ghaziabad University Courses Campus

  • Value-added diploma in data analytics: data mining, cleaning, predictive modelling and visualisation in Python.
  • President of the IT Club, class representative, Best Performer of the Batch.
  • Final project: J.A.R.V.I.S., an AI voice assistant for macOS.

Apr 2020 to Apr 2021

Intermediate, Commerce, grade A

Children’s Academy, Ghaziabad

Skills

Tools I Build With.

The languages, frameworks and platforms I use in my projects and at Siemens Healthineers.

  • Python
  • PyTorch
  • TensorFlow
  • scikit-learn
  • Hugging Face
  • pandas
  • SQL
  • Jupyter
  • Streamlit
  • SAP Signavio
  • Azure DevOps
  • Tableau
  • AWS EC2
  • OpenAI API
  • OpenCV
  • Git
  • Docker
  • Azure SQL

Based in Germany

Based in Erlangen. On-site or hybrid in Erlangen, Nuremberg, Munich, Berlin and Frankfurt. Also skilled in Copilot Studio, Power Automate, Power Apps and Power BI.

Full skill set

Generative AI and agents

Generative AI, AI agents, LLM prompting and evaluation, Microsoft Copilot Studio, Hugging Face Transformers, Qwen2.5 (open-weight LLM)

Machine learning and data science

Python, machine learning, deep learning, NLP, scikit-learn, PyTorch, model interpretability (LIME), robustness testing, forecasting

Data engineering and cloud

SQL, T-SQL stored procedures, relational data modelling, data quality and governance, Microsoft Azure (Azure SQL Server), AWS (EC2), Docker, Git

Business process and enterprise apps

SAP Signavio, interface architecture, Power Apps, Power Automate, Power BI, Excel, PowerPoint automation, Tableau

Languages: English (professional working proficiency), German (limited working proficiency), Hindi (native)

Projects

Explore, Model, and Ship.

Industry work, research and data science projects. Notebook previews are real outputs from my code.

Featured Projects

All on GitHub

2026

Ticket-to-Talk: AI agent for incident reporting, Siemens Healthineers

Copilot Studio agent that turns incident exports from the SHARP ticketing system into a filterable KPI dashboard, an incident summary and presentation-ready PowerPoint decks in the team’s template.

Industry work · Generative AI

2026

Interface Catalogue: governed data app on Azure SQL

Power Apps application on a 13-table Azure SQL data model with bulk Excel ingestion and T-SQL business rules: one source of truth for interface metadata across enterprise systems.

Industry work · Data engineering

2026

Master’s thesis: offensive language detection

Conceptual model of the thesis variables

Zero-shot and few-shot transfer with XLM-R versus in-context learning with Qwen2.5 on Kannada, Tamil and Malayalam.

NLP research

2026

Order-to-Cash IT ecosystem map, Siemens Healthineers

Modelled the Order-to-Cash interface architecture in SAP Signavio: 50+ integrated systems around the ERP as nodes and directional flows, after evaluating Power BI, LeanIX and Power Apps.

Industry work

2025

Cinemate: movie recommender on AWS

NLP pipeline on the TMDB dataset: TF-IDF vectors and a similarity matrix for content-based filtering, a Streamlit UI with custom CSS and a GPT-4 assistant, deployed on AWS EC2.

NLP app

2025

Phishing email classifier

Word clouds of phishing and legitimate emails

Text classification that separates phishing emails from legitimate ones, from cleaning and vectorising to model comparison.

NLP

2025

Walmart weekly sales forecasting

Box plot of Walmart weekly sales

Forecast weekly sales per store from holidays, temperature, fuel cost, CPI and unemployment, to support inventory and staffing decisions.

Machine learning

2025

Bangalore house price prediction

Scatter plot of price against area

Compared linear regression, decision tree, random forest and SVR, tuned with GridSearchCV and cross-validation on area, location, size and amenities.

Machine learning

2025

Titanic survival prediction

Survival counts by sex

EDA, feature engineering on class, age, sex, fare and family, and a logistic regression classifier in scikit-learn.

Machine learning

2025

Netflix catalogue analysis

Netflix content by country chart

Exploratory analysis of content growth, countries, genres, ratings, directors and cast.

Data analysis

2025

Gym member analysis

Workout intensity by gender chart

Workout type, frequency, calories and body-fat patterns across gym members.

Data analysis

2025

Rice variety image classification

A deep learning model that classifies rice grains into varieties from images.

Deep learning

2024

J.A.R.V.I.S. AI voice assistant

JARVIS assistant interface

My bachelor’s final project: a macOS voice assistant with face-recognition login, speech recognition, Hugging Face chat and image generation, and system automation. Python with OpenCV and Eel.

AI app

2024

Expense tracking system

Expense tracker screenshot

Built during IIT Kanpur winter training: tracks and groups expenses by year, month and week. HTML, CSS, JavaScript, PHP and Bootstrap.

Web

2023

QR attendance tracker

QR attendance screenshot

Records student entry and exit times. QR codes only validate through the built-in scanner, and teachers can generate codes and edit records.

Web

Research

Question, Test, and Explain.

My master’s thesis, with code and notebooks on GitHub.

MSc thesis, Gisma University of Applied Sciences, winter 2026. Supervisor: Prof. William Morrison.

Cross-lingual and few-shot offensive language detection in low-resource Dravidian languages

Social platforms need to moderate comments in languages that have almost no labelled data. I compared two cheap ways to get there: fine-tuning the multilingual transformer XLM-R with 0 to 500 target-language examples, and prompting the small LLM Qwen2.5-1.5B with in-context examples. I evaluated on code-mixed Kannada, Tamil and Malayalam, with Macro-F1, LIME explanations and three random seeds.

0.61

Macro-F1 on Kannada with zero Kannada labels, transferring from English, German and Hindi.

6 of 6

Language and shot settings where fine-tuned XLM-R beat in-context learning with Qwen2.5.

~90%

Of the full-data baseline reached with only 500 labels for Kannada and Tamil.

96%

Accuracy from a Malayalam model that predicted one class. Macro-F1 exposed it at 0.49.

Conceptual model: sample size, learning mechanism and class imbalance affect classification quality and its variance across seeds
Conceptual model of the study.

What I’d tell a team

  • Report Macro-F1, not accuracy, when classes are imbalanced.
  • Run several seeds before trusting a few-shot result.
  • Try a class-weighted simple baseline before a large model.
PyTorchHugging FaceXLM-RoBERTaQwen2.5scikit-learnLIME
Code and notebooks

Certifications

Learn, Prove, and Repeat.

Verified credentials from IBM and FutureLearn, plus certificates, awards and letters of recommendation.

Certificate collections

Verified credentials

Siemens HealthineersGisma UniversityIIT KanpurIMS GhaziabadOpen WeaverYBI FoundationTechnoHacksEduSkillsGUS Germany

Worked and learned with

Connect with me

I’m looking for data scientist, ML engineer, AI automation and data analyst roles, on-site or hybrid in Erlangen, Nuremberg, Munich, Berlin or Frankfurt.