Driven by technical excellence and architectural innovation, my research and open-source contributions focus on three core pillars that define the next generation of scalable and secure ecosystems.
AI
Active contributor to the open-source AI ecosystem, focusing on model optimization, local inference performance, and multi-modal architectures.
Automation
Designing advanced smart home infrastructures with custom integrations and privacy-focused domestic ecosystems.
Robotics
Researching autonomous navigation, sensor fusion, and real-time processing using frameworks like ROS2 for experimental hardware platforms.
=כּsפּ ג ףa£לחףוx
A selection of active research projects and experimental laboratories. These initiatives serve as a testing ground for emerging technologies, ranging from systems programming to autonomous AI orchestration. For a comprehensive overview of my professional trajectory and past technical leadership, please refer to my LinkedIn
Most of the underlying services are self-hosted on a lightweight 3-node low-power home cluster.
The HNPaper
An autonomous AI-driven news engine that aggregates, synthesizes, and renders daily tech insights. It integrates LLM orchestration for content distillation and high-fidelity TTS pipelines for automated audio generation, serving as a live testing ground for agentic content curation.
Silence Agent
A secure, local-first agentic AI platform. An autonomous AI agent that reasons, plans, and acts on its own through tool use, native voice, persistent memory, messaging, and browser automation. Built as a single Rust binary running entirely on user-owned hardware, with sandboxed execution. Positioned as a security-first alternative to agentic runtimes like OpenClaw and Hermes. [demo possible]
Hello Palo
Hello Palo helps employees take care of their mental health and better understand their difficulties, in order to prevent crises and create the conditions for a more peaceful work environment. It does so through a conversational companion grounded in the Palo Alto approach: the employee talks with 'Palo' during a framed session of about thirty minutes, then receives a synthesis of their situation, a measured intensity level, and a list of concrete actions to try. Sessions build on one another within a single journey. Palo carries forward the context of past conversations and actions, creating genuine follow-up rather than a throwaway chat. Any session can be submitted for review by a professional before its actions are revealed: the AI proposes, a human validates.
Kaio
A high-performance, low-overhead container management dashboard, CLI, and TUI. Engineered in Rust for safety and speed, Kaio provides deep observability into Docker ecosystems. It combines container observability and Docker Compose stack management in a single Rust binary running on the host. Multiple servers can also join a cluster, allowing them to discover one another, monitor their availability, and be managed from a single interface. Kaio speaks the Model Context Protocol (MCP), allowing an AI agent to connect directly to Kaio and access its state through callable tools. This means a request like “deploy nextcloud” can become an action the agent executes by itself. [Demo available]
La Plume
La Plume transforms content you have already published into a clean, durable personal website. The idea is simple: reclaim ownership of your content and make it independent from any platform or ecosystem. Powered by a single Rust binary, La Plume handles content extraction, cleanup, and Markdown generation, then builds a fast and minimalist static website using Astro. The result is a simple, portable, and maintainable architecture focused on long-term durability. [demo possible]
Dropflow
A privacy-centric, high-throughput file transfer platform built on a Zero-Trust security model. Leveraging Rust's concurrency and Axum's performance, it enables secure, large-scale data exchange via expiring cryptographic links and S3-compatible storage.
sh@tty
A localized AI assistant designed for absolute data sovereignty. By leveraging model quantization and local inference via llama.cpp, it delivers a secure LLM experience on consumer-grade hardware without compromising privacy or performance.
La Case
A unified media ingestion and management platform. It automates media retrieval tasks through an event-driven architecture, featuring real-time synchronization via SSE and a self-contained single-binary deployment for maximum portability. [demo possible]
voxify
A sophisticated media transformation engine that converts text into high-quality, synchronized audio-visual content. Optimized for multithreading, it generates precise word-level temporal metadata and VTT subtitles for seamless accessibility integration.
HNPaper
A distributed search architecture for Hacker News, capable of real-time indexing and advanced querying. Originally built as a high-performance demonstration of Elasticsearch pipelines, it showcases efficient handling of massive, live data streams.
Planning Poker
A real-time collaborative platform for agile estimation. Built with a focus on state synchronization and low-latency communication, it facilitates team consensus through a seamless, distributed card-voting system.
Levenshtein Visualizer
An interactive pedagogical tool for algorithmic logic. It visualizes dynamic programming patterns through a real-time matrix animation, illustrating the mathematical cost of string transformations in a highly intuitive manner.