Skip to content
View tes2000's full-sized avatar

Highlights

  • Pro

Block or report tes2000

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
tes2000/README.md

Hi, I'm Therese Vince Mathew 👋

Data Scientist | Machine Learning | Deep Learning | Applied AI

Based in Erlangen, Germany 🇩🇪

LinkedIn Email GitHub


👩‍💻 About Me

I am completing my M.Sc. in Data Science at Friedrich-Alexander- Universität Erlangen-Nürnberg, with a focus on Machine Learning and Artificial Intelligence.

My work covers predictive modelling, deep learning, explainable AI, data engineering and LLM-powered applications. I enjoy building complete data products—from data ingestion and preprocessing to model evaluation, interactive applications and interpretable results.

For my master's thesis, I developed and evaluated machine-learning models for used-vehicle price prediction on more than 240,000 records, comparing performance under conventional IID evaluation and realistic temporal distribution shift.

I am currently seeking full-time opportunities in Germany in:

  • Data Science
  • Machine Learning
  • Applied AI
  • Data Analytics
  • Junior ML Engineering

🚀 Featured Projects

🧭 CareerFit AI

An applied AI platform designed to analyse job requirements, extract relevant skills and identify gaps between candidate profiles and market expectations.

Key components

  • Candidate and job-profile analysis
  • Structured skill extraction and taxonomy
  • Skill-gap identification
  • FastAPI backend
  • Streamlit user interface
  • Modular Python application structure

Technologies: Python, FastAPI, Streamlit, pandas

View repository


🚗 Used-Vehicle Price Prediction Under Distribution Shift

Master's thesis comparing linear, tree-based and neural-network models for used-vehicle price prediction under IID and temporal evaluation settings.

Highlights

  • Built an end-to-end machine-learning pipeline for 240K+ vehicle records
  • Implemented Linear Regression, Random Forest, XGBoost, CatBoost and ANN models
  • Designed separate IID and temporal evaluation strategies
  • Achieved an IID test R² of 0.917 with XGBoost
  • Analysed model robustness when predicting prices for future vehicle listings
  • Compared predictive performance, generalisation and model stability

Technologies: Python, pandas, NumPy, scikit-learn, XGBoost, CatBoost, PyTorch

Repository will be published after the final thesis result and academic requirements are completed.


🤖 Automotive Data Intelligence Platform

An LLM-powered automotive analytics assistant that allows users to explore large vehicle datasets using natural-language questions.

Highlights

  • Supports arbitrary CSV uploads
  • Performs automatic schema detection and preprocessing
  • Handles missing values and numeric type conversion
  • Integrates OpenAI and Groq language models
  • Enables natural-language querying over datasets containing 100K+ records
  • Provides interactive summaries, exploration and query results
  • Includes safeguards for parsing errors and application state

Technologies: Python, pandas, Streamlit, LangChain, OpenAI API, Groq API


🌱 Efficient CNN Inference for Agricultural Image Classification

A deep-learning project focused on potato leaf disease classification and efficient model deployment across different hardware platforms.

Highlights

  • Achieved up to 97.29% classification accuracy
  • Evaluated models on T4 GPU, RTX 2080 and Coral TPU hardware
  • Compared accuracy, latency, power usage and energy efficiency
  • Applied TensorRT and TensorFlow Lite optimisation
  • Reduced measured inference time from 7.96 seconds to 0.23 seconds
  • Investigated TensorFlow retracing and graph-execution bottlenecks

Technologies: TensorFlow, Keras, TensorRT, TensorFlow Lite, CNN, Python


🔬 Additional Project Highlights

Crime and Weather Correlation Analysis — Los Angeles

Built an ETL and analytics pipeline combining Los Angeles crime and weather data from 2020–2023.

  • Integrated crime and weather datasets through a structured ETL pipeline
  • Performed temporal and geospatial analysis
  • Investigated seasonal patterns and high-density crime locations
  • Examined statistical associations between weather variables and crime rates
  • Created maps and visualisations for communicating results

Technologies: Python, pandas, PostgreSQL, Matplotlib, Folium

CNN versus Vision Transformer Image Classification

Compared VGG-16 and ViT-B/16 architectures using a multi-class image dataset containing more than 22,000 images.

  • CNN test accuracy: 90.60%
  • Vision Transformer test accuracy: 95.40%
  • Analysed convergence, architecture, model capacity and computational trade-offs
  • Compared convolution-based and attention-based feature representations

Technologies: Python, PyTorch, VGG-16, Vision Transformer

Explainable AI for Predictive Quality in Manufacturing

Developed predictive models for manufacturing quality control and used explainability methods to make model outputs more understandable.

  • Built XGBoost and LSTM models for predictive quality analysis
  • Applied SHAP to identify influential input variables
  • Created Power BI dashboards for model outputs and feature importance
  • Translated model results into decision-support information

Technologies: Python, XGBoost, LSTM, SHAP, Power BI

FAU Innovation Ecosystem Platform

Designed a user-centred platform concept to improve access to the innovation ecosystem in Fürth.

  • Conducted user and stakeholder research
  • Applied Design Thinking and the Double Diamond framework
  • Developed interactive map and recommendation concepts
  • Defined user journeys, value propositions and customer segments
  • Contributed to prototyping and business-model development

Methods: Design Thinking, User Research, Prototyping, Business Model Canvas


🛠️ Technical Stack

Programming

Python SQL R Bash C++

Machine Learning and Deep Learning

scikit-learn XGBoost PyTorch TensorFlow Keras

Data Science and Analytics

pandas NumPy PostgreSQL Power BI Tableau

Applications and Generative AI

FastAPI Streamlit LangChain OpenAI

Development Tools

Git GitHub Linux Jupyter LaTeX


🎓 Education

Degree Institution Status
M.Sc. Data Science FAU Erlangen-Nürnberg, Germany Thesis completed; final result pending
B.Sc. Computer Science, Mathematics and Statistics CHRIST University, India First Class, CGPA 9.05/10

🏅 Achievements

  • e-fellows.net Scholarship Holder — selected for a leading German career and academic network
  • Rashtrapathi Award, Bharat Scouts and Guides — national recognition for leadership and community service
  • Student Welfare Office Member, CHRIST University — supported student engagement and peer-guidance initiatives

🌍 Languages

English German Malayalam


📫 Connect With Me

I am open to full-time opportunities and collaborations involving machine learning, data science, deep learning, explainable AI and applied AI systems.

Popular repositories Loading

  1. datasciencecoursera datasciencecoursera Public

  2. datasharing datasharing Public

    Forked from jtleek/datasharing

    The Leek group guide to data sharing

  3. MADE MADE Public

    Forked from jvalue/made-template

    Template repository for the Methods of Advanced Data Engineering course at FAU

    Jupyter Notebook

  4. certificates certificates Public

  5. careerfit-ai careerfit-ai Public

    Python

  6. tes2000 tes2000 Public