Data Scientist | MSc in Data Science from BGU
I'm a Data Scientist passionate about building intelligent systems with Machine Learning and Generative AI. My background combines academic research, data science, and product-oriented thinking. I enjoy turning complex ideas into practical AI solutions and I'm currently exploring LLMs, agentic AI, RAG, and evaluation frameworks.
A simulation framework for collaborative document writing using LLM-based agents, presented at ECAI 2025.
- Problem: Explored how LLM-based agents can collaborate on shared document writing through dynamic propose-and-vote interactions.
- Architecture: Designed and implemented multi-agent simulations in Python, including agent roles, interaction workflows, and collaborative decision-making mechanisms.
- AI Workflow: Built end-to-end Generative AI workflows using OpenAI APIs, LangChain, and RAG, including prompt engineering, agent orchestration, and evaluation pipelines.
- Evaluation: Applied NLP, recommendation algorithms, and data analysis techniques to analyze large-scale simulation outputs, identify behavioral patterns, and evaluate system performance.
Enhancing HuBERT-based Speech Emotion Recognition through data augmentation, fine-tuning, and cross-dataset evaluation.
- Problem: Investigated how data augmentation improves the robustness and generalization of Speech Emotion Recognition (SER) models under diverse acoustic conditions.
- Deep Learning: Fine-tuned a pre-trained HuBERT model for seven emotion classes and evaluated SpecAugment, Time Stretch, Additive Noise, and Neutral Copy-Paste augmentation strategies.
- Evaluation: Designed a comprehensive experimental framework comparing individual and weighted augmentation combinations, including cross-dataset validation on the SAVEE corpus.
- Key Findings: Demonstrated that frequency-domain augmentations significantly improved recognition performance while highlighting the remaining challenges of cross-corpus generalization.
An end-to-end Business Intelligence platform combining data warehousing, interactive dashboards, and predictive analytics for railway operations.
- Architecture: Designed a three-layer BI architecture consisting of a Star Schema data warehouse, Tableau dashboards, and Python-based analytics.
- Data Engineering: Built a dimensional data model integrating operational, ticketing, scheduling, and station data into a unified analytical platform.
- Analytics: Applied K-Means clustering and Gradient Boosting to segment station profiles and forecast passenger demand for operational planning.
- Business Intelligence: Developed executive dashboards and OLAP reports to monitor KPIs, operational performance, revenue trends, and passenger flow.
| Category | Tools & Technologies |
|---|---|
| Languages | Python, SQL, R, Java |
| Data Analysis | Pandas, NumPy, Matplotlib, Statistical Analysis, Experimentation |
| Machine Learning | Scikit-Learn, PyTorch, TensorFlow, Model Evaluation |
| Generative AI & NLP | APIs, LangChain, LangGraph, LLMs, RAG, NLP, Transformers, Vector Databases |
| Tools | Git, Docker, Jupyter Notebook, VS Code |
| Core Strengths | Research, Problem Solving, Product Thinking, Data-Driven Decision Making |
A Dynamic Approach to Collaborative Document Writing |
Lead author • Presented at ECAI 2025, an A-ranked international AI conference.
