Optimization of an Organic Rankine Cycle

In this project, I optimize the composition of the working fluid and the operating pressure of an Organic Rankine Cycle (ORC). The optimization is conducted using genetic algorithms, specifically NSGA-II, which interacts with a custom-designed hybrid Python-Aspen Plus platform. This framework facilitates efficient data exchange and the integration of optimization algorithms with advanced process simulation.

SQL chatbot

In this project, I develop an application using LangChain that enables users to connect to a SQL database by entering its URI and posing questions in natural language. The application translates these questions into SQL queries, executes them, and returns the answers in an easy-to-read format.

Semantic search - book recomender

In this project, I develop a semantic book recommendation system that allows users to receive personalized reading suggestions based on a natural language query. The application loads a dataset of books enriched with emotional metadata and uses LangChain with Google Generative AI embeddings to semantically search a corpus of book descriptions. Users input a query, along with optional category and emotional tone filters, and the system retrieves and ranks relevant titles using a FAISS vector store. Recommendations are presented in a human-readable format, highlighting title, author(s), and a short description snippet.

Dog breed classifier

In this project, I dive into a common yet advanced topic in Deep Machine Learning: Transfer Learning. The scope of this work is to fine-tune a state-of-the-art model (VGG19) to make predictions on a dataset of over 10,000 dog images of 120 different breeds.

Conversational RAG Chatbot

This project uses generative AI and vector search to enable a chatbot that retrieves and answers questions based on content from user-provided websites, enhancing information accessibility and interaction.

Physics Informed Neural Networks (PINNs)

In this project I present a code to solve a diffussion equation (PDE) with its initial and boundary conditions using a state-of-the-art methodology PINNs

Carbon Emission calculator

This web-app estimates carbon emissions using Neural Network models. Discover how choices in diet, transport, and energy usage contribute to your carbon footprint

Hand gesture sign detector

This project uses YOLOv11 to detect and interpret hand gestures in real-time. The model has the ability to recognize gestures such as "Hello," "I Love You," "Yes," "No," and "Thanks," enhancing communication and accessibility.

House Price prediction

In this project I build a house price regressor using PyTorch. Advanced topics such as Box-Cox dependent variable transformation, PyTorch data loaders or Tabular Neural Networks (ideal to handle categorical and continuous data) are presented.

Steel plate defect classifier

This project utilizes machine learning techniques to predict defects in steel plates by building models that accurately classify defect presence based on a dataset with various features. After comprehensive exploratory data analysis (EDA) and model training, two approaches—a neural network and ensemble methods including XGBoost and LightGBM—were compared, with the neural network demonstrating superior accuracy in defect prediction.

Neural Stlye Transfer

This project leverages advanced deep learning techniques to apply the artistic style of one image to another, creating visually stunning results. The provided code is flexible, allowing users to download it and apply the style transfer to any input image and style of their choice.

Coffee Sales

In this project, I demonstrate my Excel abilities by analyzing coffee sales in three countries. I utilize tools such as LOOKUPs, dynamic tables, and pivots to construct an interactive dashboard.

Chat with my bot

Address

Ernst and Young
EY Technology, Ground floor
Mergenthalerallee 3-5
Eschborn, 65760
Germany

Around the web