Alexandru Aslău

Lists of some links for projects (for more projects please check the github Website not available. Sign in: https://www.twine.net/signup Website not available. Sign in: https://www.twine.net/signup Website not available. Sign in: https://www.twine.net/signup Website not available. Sign in: https://www.twine.net/signup Website not available. Sign in: https://www.twine.net/signup I have unpublished paper in medical image segmantation I developed a new architecture in Deep Learning the name is MADoubleResUnet++, this architecture is combination of Unet, DoubleU-Net, ResUnet and attention mechanism (such as SqeezNet and so on). I use this architecture for segmanting polyps. I have experience in python and I use tensorflow and keras, some pytorch, some pytorch lightning, sklearn, matplotlib, seaborn and so on. Programming Languages: Python, TypeScript/JavaScript, Java, SQL, PHP, C/C++, C# and Matlab (First I use almost all programming languages in Bachelor’s degree and in Master’s degree and second I use Python (Scikit-learn , TensorFlow, Langchain and many more) and TypeScript/JavaScript in personal projects, Headstarter Accelerator, non-research internships and so on. Software: Scikit-learn , TensorFlow, Langchain, JupyterNotebook, Google Colab, Kaggle Notebook, Pandas, Seaborn, Matplotlib, Numpy, Pytorch Lightning (some), Pytorch (some), Groq, Transformers, Ngrok, google-generativeai, OpenAI (with Groq Api Key), Streamlit and Gradio.

Brain Tumor Classification| Open-Source( ~ inhours)-GithubNov 2024-Nov2024• Used neural networks in Python toclassify 1000 MRI scans into 3 types of possible brain diseases with custommodel•Generated multimodal MRI reports for neurosurgeonsin under 200MS after image classification, construction &trainingCredit CardFraud Detection with ML|Open-source(~ inhours)-GithubOct 2024–Oct 2024• UsedML algorithms (e.g.XGBoost,Random Forest,K-Nearest Neighbors,SVM and so on) in Python to classify fraudulentor not, usedCredit Card Transactions Fraud Detection Dataset, Llama 3.1/3.2, Groqto evaluate accuracy of predicting•In this project, the task is to build an ML modelto determine whether or not a credit card transaction is fraudulent or not.US-Bank Churn Prediction|Open-source (~ inhours)-GithubSep 2025-Oct2025•Used 30k+ data set, Llama 3.1b, Groq and/or Vercel to evaluate accuracy of predicting when banking customer quits•Created an end-to-end solution complete with sending automated personalized email to banking customer based on featureengineering, normalization, model training, evaluating and hyperparameter tuning across 5 LLM models