CS Student @ Tel Aviv University (School of CS & AI) · Systems · ML/NLP · Software Engineering

• 🎓 Computer Science student at Tel Aviv University
• 🔬 NLP Research - fine-tuned DistilBERT for toxic language detection, co-authored academic paper
• ⚙️ Love low-level systems programming in C - shell, kernel module, TCP server
• 🤖 Built a neural network inference engine from scratch in C (258x faster than PyTorch on small networks)
• 📱 Built a full Android e-commerce app independently in high school

Languages

Systems & Infrastructure

ML / NLP

Data

Tools

• tiny-ml-runtime — Generic neural network inference engine in pure C. 258x faster than PyTorch on small networks. Lock-free atomics, SIMD-friendly memory layout, reveals BLAS optimization crossover on large networks.
• Algoscope — Real-time algospeak & toxicity detection on Bluesky. Fine-tuned DistilBERT model achieving 73.2% recall on MADOC dataset. Live Demo
• Clustering-Algorithms-Lab — Three clustering algorithms (Lloyd's K-Means, K-Means++, SymNMF) with 5.5x performance optimization. Hybrid Python-C architecture with comparative benchmarking framework.

• Linux-System-Programming — Production-grade C11: mini shell with multi-stage pipelines + thread-safe queue achieving ~400K items/sec. Demonstrates process management, signal handling (SIGINT/SIGCHLD), fork/exec/pipe, POSIX threads, condition variables, and memory-safe concurrency patterns.
• Kernel-Development-Lab — Linux kernel character device driver implementing message slots IPC mechanism. Features IOCTL command handling, user/kernel-space data transfer (copy_from_user/copy_to_user), atomic operations, lazy allocation, and comprehensive error handling.
• Network-Infrastructure-C — TCP printable-characters counter: single-threaded server/client streaming files with ASCII classification and robust error semantics. Includes integration tests, Makefile automation, and Docker deployment.

• CareerOS — End-to-end AI job search automation: CV analysis, interview prep scheduling, email classification, portfolio matching. Built with n8n workflows, Groq LLaMA, GitHub API, and Obsidian sync. Live Demo

• my-project-template — Professional GitHub template repository with language-agnostic Makefile, multi-stage Docker, GitHub Actions CI/CD, and interview-focused INTERVIEW_NOTES.md documentation.

Performance Optimization · Neural Network Inference · Clustering Algorithms · Linux Kernel Development · Character Device Drivers · IOCTL Protocol Design · TCP/IP Networking · Systems Programming · Process Management · Concurrency Patterns · Signal Handling · Memory Safety · C · Python

Proven Expertise:

• Performance Engineering: 258x NN inference speedup, 5.5x clustering optimization through low-level C and algorithmic refinement
• Kernel Programming: Character device drivers, IOCTL command handling, user/kernel-space data marshaling (copy_from_user/copy_to_user), atomic operations
• Concurrency: POSIX signals (SIGINT, SIGCHLD), C11 threads, mutexes, condition variables, per-thread handoff patterns, deadlock-free synchronization
• Production ML: Fine-tuned NLP models (73.2% recall), train/serve parity, threshold engineering
• Full-Stack Automation: Workflow orchestration (n8n), cloud API integration, real-time data pipelines
• Systems Programming: fork/exec/pipe, process groups, file descriptor handling, network protocols, robust error handling with standard errno semantics