I'm a Full-Stack Developer and independent researcher with a strong academic foundation in Computer Science, passionate about exploring AI, computational biology, and theoretical computer science. My journey into computer science stems from a lifelong curiosity about programming, which I transitioned into full-time after a successful career in post-production.

I enjoy working on projects that blend cutting-edge research with practical software engineering, and I'm eager to collaborate, learn, and contribute to impactful interdisciplinary work.

• Artificial Intelligence: Machine learning, neural networks, and search algorithms
• Computational Biology: Genetic algorithms, cellular automata, and their real-world applications in biological systems
• Theoretical Computer Science: Algorithm design, automata theory, and emergent computational paradigms
• Full-Stack Development: Building scalable, interactive, and user-friendly web applications

A multi-diet, portion-aware recipe search and transformation system. Evolved from Argmax's challenge into a full product with real-time analysis, scoring, and conversion for Keto, Vegan, or both, with dynamic substitutions and an interactive UI.

Version 3.0 is now in progress, adding a local-first RAG Chat interface with multi-turn memory, citations, and tool-use for searching, transforming, and exporting recipes through natural language.

Highlights:

• Portion-aware classification with practical compliance thresholds (e.g., #keto:0.8)
• Multi-diet logic, real-time scoring, and visual diet badges
• One-click recipe conversion via dynamic substitution engine
• Ingredient-level analysis UI and saved variations with unique ID management
• API and CLI for search, batch conversion, and export
• RAG Chat integration with local LLMs, structured tool-calls, and source-grounded answers

Keywords: Diet Classification, Recipe Transformation, Portion-Aware Analysis, Multi-Diet Logic, RAG Chat, LLM Tool Use, Data Science, Machine Learning, API, UI

Tools: Python, NumPy, Pandas, Scikit-learn, Matplotlib

A full end-to-end scientific pipeline that reconstructs exoplanet transmission spectra from simulated Ariel telescope data.

Highlights:

• Designed a multi-stage modular ML pipeline: calibration, noise reduction, transit detection, feature engineering, wavelength-wise regression, OOF cross-validation, and uncertainty modeling
• Implemented per-λ ridge regression models with systematic error handling, bootstrap resampling, and a custom SigmaPredictor for variance estimation
• Built clear separation between components (calibration → detection → regression → uncertainty), similar to agent tool-chains with deterministic execution steps
• Produced high-quality analysis, figures, and scientific storytelling aligned with NeurIPS-level expectations

Keywords: Exoplanet ML, Time-series modeling, Uncertainty estimation, Scientific pipelines, Regression per wavelength

Tools: Python, NumPy, CuPy

Built an end-to-end neural framework from scratch, without deep learning libraries.

Highlights:

• Implemented perceptrons, softmax regression, loss functions, AdaGrad, LR schedules, early stopping, and checkpointing manually
• Added GPU acceleration via CuPy for matrix operations
• Built custom diagnostics: convergence curves, gradient inspections, and model comparison utilities
• Demonstrates deep understanding of the internals behind modern ML systems and numerical optimization

Keywords: Numerical ML, Optimization algorithms, GPU computing, Educational ML framework

Tools: Python, Scikit-learn, Pandas

A complete supervised classification pipeline for medical risk prediction.

Highlights:

• Full preprocessing workflow: imputation, medical feature discretization, normalization, SMOTE for imbalance
• Compared multiple models (Random Forest, SVM, Logistic Regression, Naive Bayes, C4.5)
• Included ROC/PRC curves, confusion matrices, threshold tuning, and metric interpretation
• Emphasized reliable evaluation and model behavior analysis, relevant for constructing robust decision systems

Keywords: Classification, Healthcare analytics, Model evaluation, Imbalanced data

A genetic algorithm implementation designed to discover Methuselah patterns in Conway's Game of Life. Methuselah patterns are small initial configurations that evolve for many generations before stabilizing.

Highlights:

• Explored evolutionary approaches to optimize and analyze pattern behavior
• Integrated visualization tools to study dynamic evolution in real-time

Keywords: Genetic Algorithm, Conway's Game of Life, Methuselah Patterns, Computational Biology

A simulation of cellular automata developed as part of the Biological Computation course. This project explores emergent behaviors in systems and their applications, such as modeling environmental and climate-based models.

Highlights:

• Designed custom automata rules for advanced simulations
• Utilized Python with Tkinter for visualization and statistical analysis of system dynamics
• Investigated environmental and climate-based models through computational experiments

Keywords: Cellular Automaton, Climate Simulation, Computational Biology

This project demonstrates a simplified DNA Fountain encoding and decoding system. It encodes binary data into "droplets" (small units) using an XOR-based combination of data chunks and maps the resulting binary data to a DNA sequence using the nucleotides A, C, G, T.

Highlights:

• Implemented DNA Fountain coding for data storage in DNA
• Demonstrated encoding and decoding processes

Keywords: DNA Fountain, Data Storage, Computational Biology

A custom assembler written in C as part of the Systems Programming Lab course at the Open University of Israel. The assembler translates assembly language instructions into machine code for a specific architecture.

Highlights:

• Implemented instruction parsing and error handling for assembly programs
• Designed a robust system for macro expansion and machine code generation
• Demonstrated low-level programming expertise and a deep understanding of systems programming

Keywords: Assembler, C Programming, Systems Programming, Low-Level Programming

A single-page weather forecast application designed with modern web technologies. The app integrates React.js for the frontend, Node.js for the backend, and MongoDB for data storage, leveraging a REST API for weather data retrieval.

Highlights:

• Developed a responsive UI with React and REST API integration
• Full-stack implementation showcasing robust backend and frontend interaction

Keywords: SPA, Node.js, React.js, MongoDB, Weather Forecast API

A Chrome extension created for Open University students to calculate their GPA and plan future grades using various methods.

Highlights:

• Provided a user-friendly interface for practical academic assistance
• Designed to streamline student grade calculations and progress tracking

Keywords: Chrome Extension, Academic Tools, GPA Calculator

A responsive portfolio website built for May Vitelson, showcasing her UX/UI design with elegant and functional frontend implementation.

Highlights:

• Collaboration-focused project with UX/UI and development integration
• Frontend development optimized for user experience and responsiveness

Keywords: Portfolio Website, Frontend Development, Collaboration

• 🌱 Currently exploring the intersection of theoretical computer science and AI-driven solutions
• 💻 Passionate about bridging theory and practical application through innovative projects
• 🎯 Open to collaborations, research, and professional opportunities

• LinkedIn
• GitHub

Thank you for visiting my profile! I'm eager to connect, learn, and contribute to impactful projects in computer science and beyond. 🚀