class Engineer:
def __init__(self):
self.name = "Asaf Lecht" # אסף לכט
self.role = "Full-Stack AI/Data Engineer"
self.location = "Israel"
self.stack = ["Python", "TypeScript", "Google Cloud",
"Gemini", "Playwright", "React", "Docker"]
def mission(self):
return "Bridging legacy enterprise data with modern AI architectures."I build production-ready systems that replace manual workflows with AI-powered automation. My work spans manufacturing video analysis, document processing pipelines, RAG knowledge bases, and browser automation. I write about the process -- struggles, breakthroughs, and lessons -- on my blog. I also built Smart Hebrew RTL, a Chrome extension with 80+ users for fixing Hebrew typing in AI chats.
| Languages | AI & Data | Frameworks | DevOps & Tools |
|---|---|---|---|
    |
    |
     |
    |
AI-Powered Video Analysis for Manufacturing
- The Tech: Python (Flask + SocketIO), TypeScript (React), FFmpeg, Docker.
- The Solution: Full-stack app that analyzes manufacturing footage using LLMs. Custom frame extraction, video fingerprinting, and a modular architecture for swappable analysis services.
Digital Signage System
- The Tech: PyQt6 (desktop), Flask (web), Hebrew Calendar APIs.
- The Solution: Dual-platform (desktop + web) display for a synagogue -- prayer times, Hebrew calendar, memorial slideshows, and RTL support. Ships as a Windows installer.
Automated Document Extraction
- The Tech: Google Apps Script, AI extraction, HTML/JS dashboard.
- The Solution: PDFs arrive via email or cloud storage, AI extracts structured data, and exports in the ERP's import format. Hebrew web dashboard for review and approval.
Article Extraction Service for 17 Israeli News Sites
- The Tech: Python, Docker, Google Cloud Run.
- The Solution: A web service that extracts clean article text from Israeli news outlets. Per-site extraction strategies with fallback chains, deployed as a lightweight API.
Personal Information Hub with AI Digests
- The Tech: Python, Docker, IMAP, push notifications.
- The Solution: Monitors dozens of messaging groups and email newsletters, generates a daily AI-powered digest summarizing everything important. Runs 24/7 with automated crash recovery.
Chrome Extension -- 80+ Active Users
- The Tech: JavaScript, Chrome Extensions API, CSS injection.
- The Solution: Fixes Hebrew RTL text direction across AI chat interfaces, social media, and forums. Published on the Chrome Web Store.
Batch Audio Download with Speaker Diarization
- The Tech: Node.js, commander.js, Chrome DevTools Protocol.
- The Solution: CLI tool for batch-downloading Google Recorder audio with speaker labels. Uses accessibility tree snapshots instead of CSS selectors for reliable web app interaction.
Production ETL: 20K+ WordPress Pages to LLM-Ready Markdown
- The Tech: Python, BeautifulSoup, FastAPI, multiprocessing, WP REST API, Gemini API.
- The Solution: End-to-end pipeline that extracts WordPress content, strips 88% boilerplate, deduplicates via content hashing (caught 1,900+ dupes), preserves threaded comments, and bundles 18.5M words into RAG-ready sources. Incremental updates via REST API polling. Open source.
REST API for Web Apps Without APIs
- The Tech: Python, FastAPI, Playwright, headless Chromium.
- The Solution: Wraps web applications that lack public APIs with clean REST endpoints. Persistent browser session handles auth natively. 24h endurance tested with heartbeat monitoring.
Most repositories are private. See my blog for detailed write-ups on my projects and development process.