Welcome to my portfolio. My name is Saman Pordanesh, but you can call me Sina. Software Engineer and a fan of technology, is any shape. Here is an extension of my resume where you can explore my activities in more detail. Feel free to browse through my blogs to discover additional aspects of my work and interests.
🇮🇹 Poster presentation at MSML2025, Naples, Italy. Poster title: "Statistically Accurate Invariant Measure Informed Forecasting of QG Dynamics"
📄 We are sharing our findings on "Hilbert Neural Operators" as a preprint experimental report. Ongoing research and invite you to join us, if you are interested in the topic.
📣 Volunteered and participated in ICML 2025, Vancouver, Canada. Major update on state-of-the-art AI & LLM developements. SciML meetup and find amazing friends and colleagues.
📣 Our latest paper (My FIRST Paper 🥇), 'Hiding in Plain Sight: On the Robustness of AI-generated Code Detection,' has been accepted at DIMVA 2025, Graz, Austria.
Introduction to Hilbert Neural Operator, a new approach regarding Neural Operators, Deep Learning architectures, tailored for scientific modelling and computational physics.
Read More
Contribution report on my observations at ICML2025, around the field of AI4Science and summary about some selected posters.
Read More
Who Was Rumi?
Read More
Developed a BERT-based NLP model using modern GenAI, Fine-Tuning & MLOps practices, integrating advanced data handling, evaluation, and scalable deployment techniques to enhance performance and reproducibility.
A VSCode extension that enhances code reviews using GenAI. Integrates local and remote LLMs, focusing on user privacy and operational efficiency, transforming software development practices.
"Exploring the Capabilities of Large Language Models in Binary Reverse Engineering." Here, we delve into the fascinating world of applying advanced AI models, particularly GPT-4, to the nuanced field of reverse engineering.
In this project, we predict Falcon 9 first stage landing success to estimate launch costs. SpaceX's reusable first stages reduce costs to $62 million versus $165 million from competitors. Accurate predictions aid competitive bidding against SpaceX.
This project utilizes data science and machine learning techniques to analyze and predict Netflix viewership and IMDb ratings. The primary dataset, released by Netflix, contains around 18,000 movie titles with their viewing hours, enriched with additional data from the OMDB API.
This project uses machine learning and PySpark to predict flight arrival delays from a 2007 U.S. flight dataset, exploring models like Random Forest and Linear Regression to enhance airline efficiency and passenger experience.
